# Least Privilege Principle: every page as markdown The principle of least privilege in AWS, Azure and Google Cloud: NIST AC-6, privilege creep, examples, a policy template and the Access Debt Calculator. --- # The least privilege principle, and the calculator that sizes your clean-up Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. Least privilege means each identity holds only the access its task needs and nothing more. The Access Debt Calculator on this page takes counts from your cloud estate, such as identities with unused permissions and reachable paths, and shows how big a permission-by-permission clean-up is next to a boundary plan. The Access Debt Calculator. Default example: 2,400 identities, 1,500 with unused permissions, 120 permissions granted and 9 used per identity, 120 accounts, 25 changes a week, 6 risk classes. Arithmetic: /methodology/. ## Where the inputs come from Identities with unused permissions: AWS IAM Access Analyzer unused access findings, the Google Cloud IAM recommender, or Microsoft Entra access reviews. Permissions used: last accessed information in AWS IAM and activity analysis in the other clouds. Accounts: AWS Organizations, Azure management groups or Google Cloud folders. ## The field guide - [Principle of least privilege: definition and NIST AC-6](/principle-of-least-privilege/) - [Privilege creep in the cloud: causes and how to measure it](/privilege-creep/) - [Least privilege access: how access is granted and limited](/least-privilege-access/) - [Least privilege access control: RBAC, ABAC and boundaries](/least-privilege-access-control/) - [Least privilege policy template for cloud access](/least-privilege-policy-template/) - [The first 90 days of cloud least privilege](/first-90-days/) - [Zero trust and least privilege: what NIST SP 800-207 says](/zero-trust-least-privilege/) - [Least privilege examples in AWS, Azure and Google Cloud](/least-privilege-examples/) - [IAM least privilege in AWS, Azure and Google Cloud](/iam-least-privilege/) - [Why least privilege projects stall in the cloud](/why-least-privilege-stalls/) ## By cloud and identity type - [Least privilege for human users in AWS](/guides/aws-human-identities/) - [Least privilege for service identities in AWS](/guides/aws-service-identities/) - [Least privilege for workload identities in AWS](/guides/aws-workload-identities/) - [Least privilege for third-party identities in AWS](/guides/aws-third-party-identities/) - [Least privilege for human users in Azure](/guides/azure-human-identities/) - [Least privilege for service identities in Azure](/guides/azure-service-identities/) - [Least privilege for workload identities in Azure](/guides/azure-workload-identities/) - [Least privilege for third-party identities in Azure](/guides/azure-third-party-identities/) - [Least privilege for human users in Google Cloud](/guides/google-cloud-human-identities/) - [Least privilege for service identities in Google Cloud](/guides/google-cloud-service-identities/) - [Least privilege for workload identities in Google Cloud](/guides/google-cloud-workload-identities/) - [Least privilege for third-party identities in Google Cloud](/guides/google-cloud-third-party-identities/) ## By service - [Least privilege for Amazon S3](/services/aws-s3/) - [Least privilege for AWS Lambda](/services/aws-lambda/) - [Least privilege for Amazon EC2](/services/aws-ec2/) - [Least privilege for Amazon EKS](/services/aws-eks/) - [Least privilege for Amazon RDS](/services/aws-rds/) - [Least privilege for Amazon DynamoDB](/services/aws-dynamodb/) - [All 20 services](/services/) ## Ways to remove excess access The [routes directory](/routes/) describes 24 provider tools, right-sizing products and boundary platforms by the same fields, in alphabetical order, with no score and no order of preference. ## Sources - NIST SP 800-53 Rev. 5, AC-6 Least Privilege: https://csrc.nist.gov/projects/cprt/catalog#/cprt/framework/version/SP_800_53_5_1_1/home?element=AC-6 (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/ --- # The Access Debt Calculator: how big your least privilege clean-up is Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. The methodology page shows how the Access Debt Calculator works: its six inputs, the two plans it sizes, the assumptions it makes and the arithmetic in each step. A reader can reproduce any result by hand, change any assumption, and see where the calculator stops being a reliable guide. The calculator takes seven inputs: the number of identities, the number of identities with unused permissions, the permissions granted per identity, the permissions used per identity, the number of accounts, the changes your team ships per week, and the number of risk classes a boundary plan should cover. Most of them can be read from the provider reports described on the privilege creep page, such as unused access findings. It outputs the size of two plans. The permission-by-permission plan assumes one reviewed change per identity with unused permissions, since each identity has its own policy, owner and test, and divides that by your weekly rate. The boundary plan assumes one boundary per risk class, attached high in the hierarchy and rolled out in stages over your accounts. The stage count is an assumption stated under the calculator, and you can change every input. Read the result as a ratio. If the permission plan is a hundred times the boundary plan, hand edits will not finish before the estate changes again. The calculator does not account for how long a change takes to test, how many owners must approve, or whether a boundary breaks a rare dependency. Those costs are real, and the pages on boundaries and on the first 90 days explain how to test for them. ## Every assumption, with the default example | Step | Rule | Default example | | --- | --- | --- | | Granted | Identities times the mean permissions granted per identity | 2,400 times 120 = 288,000 | | Used | Identities times the mean permissions used in the last 90 days | 2,400 times 9 = 21,600 | | Unused share | (granted minus used) divided by granted, per identity | 93% | | Permission-by-permission plan | One reviewed change per identity with unused permissions | 1,500 changes | | Weeks, permission plan | Changes divided by changes shipped per week, rounded up | 60 weeks at 25 a week | | Rollout stages | 1 plus the base-4 logarithm of the accounts, rounded up | 5 stages for 120 accounts | | Boundary plan | Risk classes times rollout stages | 6 times 5 = 30 changes | | Weeks, boundary plan | Boundary changes divided by changes shipped per week, rounded up | 2 weeks | Default example as on the home page; edition 16 Nov 2026. ## Twenty known cases The calculator is checked against twenty cases written before it was built, kept in the site's tests folder; the build fails if one gives a different answer. The table shows the first five. | Case | Inputs | Permission plan | Boundary plan | | --- | --- | --- | --- | | default example | identities 2400, unused 1500, granted 120, used 9, accounts 120, rate 25, classes 6 | 1500 changes, 60 weeks | 30 changes, 2 weeks | | one account | identities 50, unused 10, granted 40, used 4, accounts 1, rate 5, classes 3 | 10 changes, 2 weeks | 3 changes, 1 week | | four accounts | identities 200, unused 100, granted 60, used 6, accounts 4, rate 10, classes 4 | 100 changes, 10 weeks | 8 changes, 1 week | | five accounts | identities 200, unused 100, granted 60, used 6, accounts 5, rate 10, classes 4 | 100 changes, 10 weeks | 12 changes, 2 weeks | | sixteen accounts | identities 300, unused 150, granted 80, used 10, accounts 16, rate 20, classes 5 | 150 changes, 8 weeks | 15 changes, 1 week | ## Cite this Cite this line: 'Least Privilege Principle, The Access Debt Calculator, edition 16 Nov 2026, https://leastprivilegeprinciple.com/methodology/' ## Related pages - [30 questions about least privilege in the cloud](/faq/) - [The Access Debt Calculator: the data, downloadable](/datasets/) - [About](/about/) ## Sources - AWS Organizations: service control policies: https://docs.aws.amazon.com/organizations/latest/userguide/orgs_manage_policies_scps.html (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/methodology/ --- # The principle of least privilege gives every identity only the access its task needs Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. The principle of least privilege means each user, process or workload gets only the access its task needs and nothing more. NIST SP 800-53 Rev. 5 control AC-6, read on 16 Nov 2026, requires it for users and for processes acting on their behalf, and AWS, Azure and Google Cloud each publish guidance for applying it. ## NIST AC-6 states the requirement in one sentence The source text is control AC-6, Least Privilege, in NIST SP 800-53 Rev. 5. NIST publishes it in the public domain. The control statement reads, verbatim: "Employ the principle of least privilege, allowing only authorized accesses for users (or processes acting on behalf of users) that are necessary to accomplish assigned organizational tasks." The catalog entry is on the [NIST CSRC catalog page for AC-6](https://csrc.nist.gov/projects/cprt/catalog#/cprt/framework/version/SP_800_53_5_1_1/home?element=AC-6), read on 16 Nov 2026. NIST's glossary gives the shorter definition, citing the same publication: the principle that a security architecture is designed so that each entity is granted the minimum system resources and authorizations that the entity needs to perform its function ([NIST CSRC glossary](https://csrc.nist.gov/glossary/term/least_privilege), read 16 Nov 2026). The wording covers people, services and processes alike, which matters in the cloud, where most principals are not people. ## The guidance paragraph adds duties, processes and roles The AC-6 guidance says organizations employ least privilege for specific duties and systems, and apply it to system processes so that they operate at privilege levels no higher than necessary. It also says organizations consider creating additional processes, roles and accounts as necessary to achieve least privilege, and that they apply it to the development, implementation and operation of systems. In cloud terms, that is the case for one narrow role per workload instead of one shared role for a team. It also means the pipeline that deploys a workload is in scope, since the pipeline holds an identity too. A policy document alone does not meet the control. The assessment objective is that the principle "is employed", which an assessor tests in configuration and records. ## Ten enhancements turn the principle into named practices AC-6 has ten enhancements in the Rev. 5 catalog. Several map directly to cloud work, and the table lists the ones a cloud program meets first. | Enhancement | Title in the catalog | Cloud practice it points to | | --- | --- | --- | | AC-6(1) | Authorize Access to Security Functions | Limit who can change IAM, logging and key settings | | AC-6(5) | Privileged Accounts | Restrict administrator roles to named people or roles | | AC-6(7) | Review of User Privileges | Scheduled access reviews with a recorded result | | AC-6(9) | Log Use of Privileged Functions | Audit logging of administrator actions | | AC-6(10) | Prohibit Non-privileged Users from Executing Privileged Functions | Deny rules that hold whatever roles are granted | ## AWS, Azure and Google Cloud give the same instruction in their own words AWS's IAM guidance tells customers to grant only the permissions required to perform a task, by defining the actions allowed on specific resources under specific conditions. It adds that starting with AWS managed policies might not give least privilege for your use case, that Access Analyzer can generate a policy from CloudTrail activity, and that you should regularly review and remove unused users, roles, permissions, policies and credentials ([AWS IAM best practices](https://docs.aws.amazon.com/IAM/latest/UserGuide/best-practices.html), read 16 Nov 2026). Microsoft's Azure RBAC guidance says to grant users the least privilege to get their work done and to avoid assigning broader roles at broader scopes even if that initially seems more convenient. Google Cloud says basic roles include thousands of permissions across all services, so in production you should grant the most limited predefined or custom roles that meet your needs ([Google Cloud, using IAM securely](https://cloud.google.com/iam/docs/using-iam-securely), read 16 Nov 2026). ## The cloud case differs in scale and in who holds the access In a data center, a privileged account was usually a person with a ticket. In the cloud, most principals are workloads: roles, service accounts and managed identities created by pipelines, often with broad permissions granted while a developer was exploring. AWS says as much in its own advice, that you might start broad and reduce as the use case matures. Creation is fast and removal is slow, so the gap grows by default. The principle is unchanged and the work is different: measure use against grants, test each change before it ships, and set boundaries that remove whole classes of access. [Why least privilege stalls](/why-least-privilege-stalls/) covers where that work stops, and [what a boundary is](/what-a-boundary-is/) covers the third part. To size the gap in your own estate, the [Access Debt Calculator](/) turns four counts into a clean-up estimate. The [method page](/methodology/) states its assumptions, and the [first 90 days](/first-90-days/) turns the estimate into a plan. ## An auditor reads AC-6 as a list of records to produce The AC-6 assessment objects name the items an assessor examines: the access control policy, procedures for least privilege, the list of assigned authorizations, system configuration settings and audit records. A program that keeps its policies as code, its change approvals and its review results in one place produces these as a by-product of the work. The records answer two questions an assessor asks of any access: who approved it, and when it was last reviewed. A [written least privilege policy](/least-privilege-policy-template/) supplies the first document on that list. ## Related pages - [Privilege creep in the cloud: causes and how to measure it](/privilege-creep/) - [Least privilege access: how access is granted and limited](/least-privilege-access/) - [Least privilege access control: RBAC, ABAC and boundaries](/least-privilege-access-control/) ## Sources - NIST SP 800-53 Rev. 5 catalog (OSCAL JSON), control AC-6: https://raw.githubusercontent.com/usnistgov/oscal-content/main/nist.gov/SP800-53/rev5/json/NIST_SP-800-53_rev5_catalog.json (read 16 Nov 2026) - NIST CSRC, SP 800-53 Rev. 5.1.1 catalog, control AC-6: https://csrc.nist.gov/projects/cprt/catalog#/cprt/framework/version/SP_800_53_5_1_1/home?element=AC-6 (read 16 Nov 2026) - NIST CSRC glossary, least privilege: https://csrc.nist.gov/glossary/term/least_privilege (read 16 Nov 2026) - AWS, security best practices in IAM: https://docs.aws.amazon.com/IAM/latest/UserGuide/best-practices.html (read 16 Nov 2026) - Microsoft, best practices for Azure RBAC: https://learn.microsoft.com/en-us/azure/role-based-access-control/best-practices (read 16 Nov 2026) - Google Cloud, using IAM securely: https://cloud.google.com/iam/docs/using-iam-securely (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/principle-of-least-privilege/ --- # Privilege creep is access that outlives the task it was granted for Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. Privilege creep is the build-up of access an identity no longer needs, because grants are added for each new task and rarely removed. In the cloud it shows up as old roles, temporary grants left in place and unused keys. AWS, Google Cloud and Microsoft each publish a way to measure it, read on 16 Nov 2026. ## Privilege creep is the gap between what an identity holds and what it uses The term comes from people who change jobs and keep the access of every earlier job. In the cloud the same pattern applies to people, service accounts, roles and managed identities. Each task adds a grant. Nothing in the normal workflow removes one, so the total only rises. NIST control AC-6(7), Review of User Privileges, names the counter-measure: review the privileges assigned to defined roles or classes of users to validate the need for such privileges ([NIST SP 800-53 Rev. 5 catalog](https://raw.githubusercontent.com/usnistgov/oscal-content/main/nist.gov/SP800-53/rev5/json/NIST_SP-800-53_rev5_catalog.json), read 16 Nov 2026). Privilege creep is what that review exists to find. The parent requirement is on [the principle of least privilege](/principle-of-least-privilege/) page. ## Four cloud habits produce most of it Role changes come first. A person moves from one team to another and keeps the group memberships of the old team, because the new team's request is the only ticket anyone files. Project ends are second: the contractor group, the migration role and the test account stay after the work is done. Temporary grants left in place are third. An engineer receives administrator access for an incident, and the expiry is a reminder in someone's calendar. Broad starting grants are fourth. AWS says in its IAM guidance that you might start with AWS managed policies, which may not give least privilege for your use case, and then work toward it. The second half of that sequence is the one that gets skipped. - Role change: old group memberships kept after a move - Project end: roles and accounts left after the work stops - Temporary access: a grant with no expiry or no one tracking it - Broad start: a wide managed policy that is never narrowed ## Provider tools measure it from usage records Each provider reports what was granted against what was used. The reports differ in window and in what they cover, so a clean result from one is not a clean result from another. | Provider tool | What it reports | Window stated in the documentation | | --- | --- | --- | | AWS last accessed information | Services and management actions last used by an IAM identity, and by an Organizations entity | Tracking period for service information of at least 400 days | | AWS IAM Access Analyzer, unused access | Unused roles, unused IAM user access keys and passwords, unused permissions | Set when the analyzer is created | | Google Cloud role recommendations | Roles to remove or replace, based on permissions used | Last 90 days by default | | Microsoft Defender for Cloud permissions management | Excessive, unused or misconfigured permissions for users, service principals, managed identities and service accounts | Not stated on the page read | ## Read the window before you act on a finding The AWS page on last accessed information says the tracking period for service information is at least 400 days, and that action information started on different dates by service: April 2020 for Amazon S3 actions, April 2021 for Amazon EC2, IAM and Lambda actions, and May 2023 for all other services. An action that shows no activity may simply predate the tracking. Google documents that the IAM recommender looks at permissions used in the last 90 days and, by default, needs 90 days of data. For project-level recommendations the minimum can be set to 30 or 60 days, and Google says accuracy might be affected. A role used once a year, for example for failover, can sit outside both windows. List such identities by hand before any removal. ## Measure creep as a count that falls over time Four counts are enough to track it and any team can produce them without a purchased product: identities with a last-used date older than your window, access keys unused for the whole window, role assignments past their expiry, and exceptions open past their expiry date. Record the counts each quarter. Creep is a rate as much as a stock, so the useful figure is how many grants were added and how many were removed in the period. If additions outpace removals, a one-time clean-up will not hold. The [Access Debt Calculator](/) sizes the clean-up from the counts, and [why least privilege stalls](/why-least-privilege-stalls/) explains what usually interrupts it. Microsoft's Entra access reviews are the scheduled counter-measure: they review group memberships, application access and role assignments on a recurring basis so that only the right people keep access ([Microsoft Learn](https://learn.microsoft.com/en-us/entra/id-governance/access-reviews-overview), read 16 Nov 2026). The [first 90 days](/first-90-days/) places that review in the plan. ## Related pages - [Least privilege access: how access is granted and limited](/least-privilege-access/) - [Least privilege access control: RBAC, ABAC and boundaries](/least-privilege-access-control/) - [Least privilege policy template for cloud access](/least-privilege-policy-template/) ## Sources - NIST SP 800-53 Rev. 5 catalog (OSCAL JSON), control AC-6: https://raw.githubusercontent.com/usnistgov/oscal-content/main/nist.gov/SP800-53/rev5/json/NIST_SP-800-53_rev5_catalog.json (read 16 Nov 2026) - AWS, security best practices in IAM: https://docs.aws.amazon.com/IAM/latest/UserGuide/best-practices.html (read 16 Nov 2026) - AWS, refine permissions using last accessed information: https://docs.aws.amazon.com/IAM/latest/UserGuide/access_policies_last-accessed.html (read 16 Nov 2026) - AWS, IAM Access Analyzer findings: https://docs.aws.amazon.com/IAM/latest/UserGuide/access-analyzer-findings.html (read 16 Nov 2026) - Google Cloud, role recommendations overview: https://cloud.google.com/policy-intelligence/docs/role-recommendations-overview (read 16 Nov 2026) - Microsoft, permissions management in Defender for Cloud: https://learn.microsoft.com/en-us/azure/defender-for-cloud/permissions-management (read 16 Nov 2026) - Microsoft, access reviews in Microsoft Entra ID: https://learn.microsoft.com/en-us/entra/id-governance/access-reviews-overview (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/privilege-creep/ --- # Least privilege access is granted narrow, in time, and checked against use Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. Least privilege access means an identity can reach only the resources and actions its work needs, for only as long as it needs them. In cloud accounts that takes four steps: grant a narrow role, scope it to resources, set an end date, and compare it with usage records. Provider guidance read on 16 Nov 2026. ## Access is least privilege when it is narrow on four axes An access grant has four properties that can be too wide: the actions allowed, the resources they apply to, the conditions under which they apply, and the time they last. Least privilege access narrows all four. Narrowing only the actions while leaving the resource as a wildcard removes less than it appears to. AWS words the first three as defining the actions allowed on specific resources under specific conditions ([AWS IAM best practices](https://docs.aws.amazon.com/IAM/latest/UserGuide/best-practices.html), read 16 Nov 2026). The fourth is covered by time-bound assignment, which Microsoft documents for Privileged Identity Management. The mechanisms behind each are on [least privilege access control](/least-privilege-access-control/). ## Start from the narrowest role that works AWS says starting with AWS managed policies might not give least privilege for your use case, and it provides managed policies for job functions that align to common roles in an organization. Google Cloud says basic roles include thousands of permissions across all services and should not be granted in production unless there is no alternative. Microsoft advises avoiding broader roles at broader scopes, and says that when a privileged administrator role must be assigned, a narrow scope such as a resource group or resource is better than a management group or subscription. The common shape is the same: the role is the first limit and the scope is the second. Neither limit helps if the grant is then given to a shared group that dozens of people can join, so check group membership as part of the same review. ## Set an end date where the task has one Microsoft documents Privileged Identity Management as a model in which users receive temporary permissions to perform privileged tasks, so that access is granted only when users need it. Assignments can be given start and end dates, and an eligible assignment must be activated before it takes effect. A grant with an end date removes itself. A grant without one depends on a person remembering to review it, which is the pattern behind [privilege creep](/privilege-creep/). For tasks with no natural end, such as a service account for a running workload, the review cycle takes the place of the end date. ## Compare the grant with use, then remove the difference Usage records show which part of a grant is dormant. AWS Access Analyzer can generate a policy from CloudTrail activity, with a look-back of up to 90 days, and its unused access findings list unused roles, unused access keys and passwords, and unused permissions. Google's role recommendations compare a principal's permissions with those used in the last 90 days. Neither provider applies the change for you. Google states that the IAM recommender does not apply recommendations automatically, and that you apply or dismiss each one. The removal step is therefore a decision with an owner, which is where projects usually slow down. [The first 90 days](/first-90-days/) gives that step a place in a schedule. ## Workloads need the same treatment as people Workload identities are most of the principals in a cloud account. Google recommends dedicated service accounts for each application, and AWS says workloads on its compute services should use the temporary credentials of an IAM role. Microsoft's managed identities remove the need to manage credentials for a workload on Azure. The per-cloud detail is in the identity guides, for example [AWS human identities](/guides/aws-human-identities/) and [Azure workload identities](/guides/azure-workload-identities/). A short checklist for the whole path is the [written policy template](/least-privilege-policy-template/). A practical test for any single grant is to ask four questions: which actions does it allow, on which resources, under which conditions, and until when. A grant that answers all four in specific terms is least privilege access. A grant that answers "any", "any", "always" and "never expires" is not, however clean the role name looks. Run the test on the five most powerful roles in each account first, since the reduction there is the largest, and write down the answers so the next reviewer can see what was decided and why. ## Related pages - [Least privilege access control: RBAC, ABAC and boundaries](/least-privilege-access-control/) - [Least privilege policy template for cloud access](/least-privilege-policy-template/) - [The first 90 days of cloud least privilege](/first-90-days/) ## Sources - AWS, security best practices in IAM: https://docs.aws.amazon.com/IAM/latest/UserGuide/best-practices.html (read 16 Nov 2026) - AWS, managed policies for job functions: https://docs.aws.amazon.com/IAM/latest/UserGuide/access_policies_job-functions.html (read 16 Nov 2026) - Google Cloud, using IAM securely: https://cloud.google.com/iam/docs/using-iam-securely (read 16 Nov 2026) - Microsoft, best practices for Azure RBAC: https://learn.microsoft.com/en-us/azure/role-based-access-control/best-practices (read 16 Nov 2026) - Microsoft, what is Privileged Identity Management: https://learn.microsoft.com/en-us/entra/id-governance/privileged-identity-management/pim-configure (read 16 Nov 2026) - AWS, generate policies based on access activity: https://docs.aws.amazon.com/IAM/latest/UserGuide/access-analyzer-policy-generation.html (read 16 Nov 2026) - AWS, IAM Access Analyzer findings: https://docs.aws.amazon.com/IAM/latest/UserGuide/access-analyzer-findings.html (read 16 Nov 2026) - Google Cloud, role recommendations overview: https://cloud.google.com/policy-intelligence/docs/role-recommendations-overview (read 16 Nov 2026) - Google Cloud, best practices for service accounts: https://cloud.google.com/iam/docs/best-practices-service-accounts (read 16 Nov 2026) - Microsoft, managed identities for Azure resources: https://learn.microsoft.com/en-us/entra/identity/managed-identities-azure-resources/overview (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/least-privilege-access/ --- # Least privilege access control combines roles, attributes, conditions and ceilings Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. Least privilege access control is the set of mechanisms that enforce least privilege: role-based access control, attribute-based access control, policy conditions and boundaries that cap what grants can allow. AWS, Azure and Google Cloud each document all four, under different names, as read on 16 Nov 2026. ## Four mechanisms do different jobs Role-based access control bundles permissions into roles and assigns the role to an identity at a scope. Attribute-based access control decides by attributes of the identity, the resource and the request. Conditions add a test to one grant. Boundaries sit above grants and cap what they can allow. A program that uses only roles leaves three limits unused. This page covers the mechanisms. How access is granted and shrunk day to day is on [least privilege access](/least-privilege-access/), and the boundary concept in detail is on [what a boundary is](/what-a-boundary-is/). | Mechanism | AWS | Azure | Google Cloud | | --- | --- | --- | --- | | Roles | IAM roles and managed policies | Azure RBAC role assignments at a scope | IAM roles bound in allow policies | | Attributes | ABAC with tags | Azure ABAC role assignment conditions | IAM Conditions on resource and principal attributes | | Conditions on a grant | Condition element in a policy | Role assignment conditions | Conditions in role bindings and deny rules | | Ceiling above grants | Permissions boundaries, service control policies | Deny assignments | Deny policies, Principal Access Boundary policies | ## Roles are the first limit and need a scope Microsoft describes Azure RBAC as an authorization system for who has access to Azure resources, what they can do with them and what areas they have access to. The area is the scope. Microsoft's guidance is to avoid assigning broader roles at broader scopes, since the scope limits what is at risk if the security principal is compromised. AWS offers managed policies for job functions that align to common roles, and notes they include filters that limit access. Google separates basic roles, which include thousands of permissions, from predefined and custom roles, which are narrower. In all three clouds a role that is too wide cannot be corrected by scope alone, and a scope that is too wide cannot be corrected by the role alone. ## Attributes and conditions cut the number of grants AWS describes attribute-based access control as an authorization strategy that defines permissions based on attributes, and it can grant or revoke access in real time as attributes change. Microsoft says Azure ABAC builds on Azure RBAC by adding role assignment conditions, and that conditions can reduce the number of role assignments, which are limited per subscription. Microsoft's own example is a condition that lets a person read blobs only if the blobs are tagged Project=Blue. Google IAM Conditions are specified in the role bindings of an allow policy and in the rules of a deny policy, and can test resource or principal attributes. The benefit is fewer, broader-looking roles that stay narrow in effect. The cost is that a tag becomes a security control, so who can edit tags becomes a question to answer. ## Boundaries cap grants without editing them AWS permissions boundaries set the maximum permissions an identity-based policy can grant to a user or role. AWS service control policies set permission guardrails across the accounts of an organization, and grant nothing themselves. Azure deny assignments block listed actions for a principal at a scope even if a role assignment grants them. Google deny policies prevent principals from using certain permissions regardless of the roles they hold. Google Principal Access Boundary policies define the resources a set of principals can access, and attach to principal sets. These are the mechanisms that remove a class of access in one change. Because they act on many identities at once, they need a test before release, for example with the [AWS policy simulator](https://docs.aws.amazon.com/IAM/latest/UserGuide/access_policies_testing-policies.html) or Google's Policy Simulator. ## The evaluation rule decides what a combination allows AWS documents that, in policy evaluation, an explicit deny in any applicable policy overrides an allow, and that boundaries and service control policies must allow an action as well as the identity policy. Each cloud has its own order and its own exceptions, so a boundary written for one cloud cannot be read as if it worked in another. Read the evaluation page for each cloud before writing a ceiling. The per-cloud identity guides, for example [Google Cloud service identities](/guides/google-cloud-service-identities/), apply these rules to one identity type, and the [glossary](/glossary/) defines the terms used here. ## Related pages - [Least privilege policy template for cloud access](/least-privilege-policy-template/) - [The first 90 days of cloud least privilege](/first-90-days/) - [Zero trust and least privilege: what NIST SP 800-207 says](/zero-trust-least-privilege/) ## Sources - AWS, permissions boundaries for IAM entities: https://docs.aws.amazon.com/IAM/latest/UserGuide/access_policies_boundaries.html (read 16 Nov 2026) - AWS, service control policies: https://docs.aws.amazon.com/organizations/latest/userguide/orgs_manage_policies_scps.html (read 16 Nov 2026) - AWS, attribute-based access control for AWS: https://docs.aws.amazon.com/IAM/latest/UserGuide/introduction_attribute-based-access-control.html (read 16 Nov 2026) - AWS, policy evaluation logic: https://docs.aws.amazon.com/IAM/latest/UserGuide/reference_policies_evaluation-logic.html (read 16 Nov 2026) - AWS, managed policies for job functions: https://docs.aws.amazon.com/IAM/latest/UserGuide/access_policies_job-functions.html (read 16 Nov 2026) - Microsoft, what is Azure role-based access control: https://learn.microsoft.com/en-us/azure/role-based-access-control/overview (read 16 Nov 2026) - Microsoft, what is Azure attribute-based access control: https://learn.microsoft.com/en-us/azure/role-based-access-control/conditions-overview (read 16 Nov 2026) - Microsoft, list Azure deny assignments: https://learn.microsoft.com/en-us/azure/role-based-access-control/deny-assignments (read 16 Nov 2026) - Microsoft, best practices for Azure RBAC: https://learn.microsoft.com/en-us/azure/role-based-access-control/best-practices (read 16 Nov 2026) - Google Cloud, overview of IAM Conditions: https://cloud.google.com/iam/docs/conditions-overview (read 16 Nov 2026) - Google Cloud, deny policies: https://cloud.google.com/iam/docs/deny-overview (read 16 Nov 2026) - Google Cloud, Principal Access Boundary policies: https://cloud.google.com/iam/docs/principal-access-boundary-policies (read 16 Nov 2026) - Google Cloud, using IAM securely: https://cloud.google.com/iam/docs/using-iam-securely (read 16 Nov 2026) - AWS, testing IAM policies with the policy simulator: https://docs.aws.amazon.com/IAM/latest/UserGuide/access_policies_testing-policies.html (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/least-privilege-access-control/ --- # A least privilege policy template you can adopt and edit Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. A written least privilege policy for cloud environments, with [bracketed placeholders] for your organization's names, periods and owners. It covers purpose, scope, roles, rules, reviews, exceptions with an expiry date, and enforcement, and maps to NIST SP 800-53 Rev. 5 control AC-6. Template edition 16 Nov 2026. ## 1. Purpose This policy sets how [Organization] grants, reviews and removes access to its cloud environments so that every identity holds only the access its function needs. It implements the least privilege requirement of [framework, for example NIST SP 800-53 Rev. 5 control AC-6] and supports [other obligations, for example the access control requirements in your contracts]. The policy owner is [Role, for example Head of Cloud Security]. It was approved by [Approver] on [Date] and is reviewed at least every [12] months. ## 2. Scope This policy applies to every cloud account, subscription and project that [Organization] owns or operates, including [AWS accounts], [Azure subscriptions] and [Google Cloud projects]. It covers every identity: employees, contractors, partners, service accounts, roles, managed identities, pipelines and automation. It applies to production and to non-production environments that hold [production data or credentials that reach production]. Systems outside the scope are listed in [Register name], each with a named owner and a reason. ## 3. Roles and responsibilities The access owner for each account, subscription or project is named in [Register name]. The access owner approves grants, performs the reviews in section 6, and signs exceptions. The security team maintains this policy, the boundaries in section 5 and the evidence described in section 8. Engineering teams request access through [Ticketing or pull request process] and do not grant access to themselves. The internal audit function, [Function], tests this policy at least [annually] and reports to [Governance body]. ## 4. Rules for granting access Access is granted for a stated task, to a named identity, at the narrowest scope that works. Each grant records the requester, the approver, the reason and the date. These rules apply to all identities: - No identity receives a basic or administrator role in a production environment unless the access owner approves it in writing for a stated task. - People use federated sign-in with temporary credentials. Long-lived access keys are issued only under an exception in section 7. - Each workload has its own identity. Identities are not shared between workloads or between environments. - Roles are assigned to groups, not to individuals, except for [named break-glass accounts]. - Privileged roles are assigned as time-bound or on-request access with a maximum duration of [8 hours]. - Wildcards in actions or resources require a recorded reason and are reviewed in section 6. - Access is defined in code and changed by pull request, so each change has an author, a reviewer and a date. ## 5. Boundaries The security team maintains organization-level limits that hold whatever roles are granted: [the cloud regions in use], [the services in use], a block on leaving the organization with data, and protection of logging and of the boundaries themselves. A new boundary is tested against [30 days] of recent access, released first in audit or log-only mode where the cloud offers it, and enforced after the affected owners have confirmed. The previous version is kept for rollback. Changes to boundaries need approval from [Role]. ## 6. Reviews Access owners review every human and workload identity every [90 days] for privileged roles and every [180 days] for all others. A review compares what is granted with what was used in the last [90] days and records the decision to keep, narrow or remove each grant. Dormant identities with no use for [90] days are disabled and removed after [30] more days unless the access owner records a reason. The security team reports each quarter: reviews completed on time, grants removed, exceptions open, and exceptions past expiry. ## 7. Exceptions An exception is a written, approved departure from section 4. It names the identity, the access, the reason, the compensating control, the approver and an expiry date no later than [90] days after approval. Expired exceptions are removed automatically or by the access owner within [5] working days. An exception can be renewed once by the same approver and then needs approval from [Senior role]. The register of exceptions is part of the evidence in section 8. ## 8. Evidence and enforcement The security team keeps for [7] years: this policy and its versions, the list of identities and their grants, review results, approvals, exception records, boundary changes and audit logs of privileged actions. Evidence is stored in [System] with access limited to [Roles]. A breach of this policy is handled under [Disciplinary or contract procedure]. Access that conflicts with this policy and has no exception is removed by the security team after notice to the access owner, or immediately where there is an active risk. The policy is checked against [NIST SP 800-53 Rev. 5 AC-6] on each annual review. Before adopting the template, read [the principle of least privilege](/principle-of-least-privilege/) for the control it implements, [the first 90 days](/first-90-days/) for a schedule to put it into practice, and the [methodology](/methodology/) for the Access Debt Calculator that sizes the work. ## Download - [Least privilege policy template (.docx)](/downloads/least-privilege-policy-template.docx): Word document made from this page, edition 16 Nov 2026 ## Related pages - [The first 90 days of cloud least privilege](/first-90-days/) - [Zero trust and least privilege: what NIST SP 800-207 says](/zero-trust-least-privilege/) - [Least privilege examples in AWS, Azure and Google Cloud](/least-privilege-examples/) ## Sources - NIST SP 800-53 Rev. 5 catalog (OSCAL JSON), control AC-6: https://raw.githubusercontent.com/usnistgov/oscal-content/main/nist.gov/SP800-53/rev5/json/NIST_SP-800-53_rev5_catalog.json (read 16 Nov 2026) - NIST CSRC, SP 800-53 Rev. 5.1.1 catalog, control AC-6: https://csrc.nist.gov/projects/cprt/catalog#/cprt/framework/version/SP_800_53_5_1_1/home?element=AC-6 (read 16 Nov 2026) - AWS, security best practices in IAM: https://docs.aws.amazon.com/IAM/latest/UserGuide/best-practices.html (read 16 Nov 2026) - Google Cloud, using IAM securely: https://cloud.google.com/iam/docs/using-iam-securely (read 16 Nov 2026) - Microsoft, what is Privileged Identity Management: https://learn.microsoft.com/en-us/entra/id-governance/privileged-identity-management/pim-configure (read 16 Nov 2026) - Microsoft, access reviews in Microsoft Entra ID: https://learn.microsoft.com/en-us/entra/id-governance/access-reviews-overview (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/least-privilege-policy-template/ --- # The first 90 days should end with one boundary enforced, not a longer list of findings Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. A first quarter of cloud least privilege work fits the 90-day usage windows provider tools use. Spend month one on logging, data and owners, month two on choosing and simulating one class of access, and month three on audit mode, enforcement and a standing review. Finish with one boundary live and proved. ## Plan around the 90-day window The tools that propose removals from usage want about 90 days of history. Google documents a default minimum observation period of 90 days for role recommendations, adjustable to 30 or 60 days for project-level ones. AWS lets Access Analyzer read up to 90 days of CloudTrail events when it generates a policy. If logging is not on today, the first job is turning it on, because every later step depends on the data it produces. That makes a 90-day program natural. Month one collects and assigns. Month two chooses and tests. Month three enforces and sets the review that keeps the result. A program that tries to enforce in month one is guessing, and one with no enforcement by month three has produced a report. To size the whole clean-up before you plan it, use the [Access Debt Calculator](/) and read its [assumptions](/methodology/). ## Month one: inventory, logs and owners Confirm that audit logging covers every account, subscription and project, and that it is retained for at least the window you plan to use. Turn on the provider analyzers in report mode: Access Analyzer unused access findings, the Google IAM recommender, and Defender for Cloud permissions management if you run Azure. Produce a first list of unused roles, access keys and dormant identities. Assign an owner to each account or project. Removal fails when the person who can approve it cannot be found. NIST SP 800-53 control AC-6(7) calls for a review of user privileges, so record who performs that review, how often, and where the result is kept. Keep the list short enough to act on, since a first list of two thousand items teaches nothing. The [policy template](/least-privilege-policy-template/) gives the owner and review wording. ## Month two: choose one class and simulate it Pick one class of access to remove, using the data from month one. Good first candidates are the ones with few legitimate users: unused regions, unused services, access keys unused for the whole window, and roles that trust accounts outside the organization. Choose the one where you can name every principal that currently uses it. Starting small is a choice made on purpose, since the first change builds trust in the process before a larger one needs it. Write the change as code. Replay recent access under it with the AWS policy simulator, Google's Policy Simulator, or the simulation of any product you use. Review each changed outcome with its owner. A change nobody can explain stays in the branch. Keep a list of the principals that used the access during the window so you can contact each owner before enforcement. The class you pick is a boundary in the sense of [what a boundary is](/what-a-boundary-is/). ## Month three: audit mode, enforcement and the review Release the change in audit or log-only mode where the cloud offers one, and watch blocked attempts across a full business cycle. Fix or exempt each hit with a recorded reason. Then enforce, and keep the previous version ready to restore. Write the date and the evidence into the change record, because an auditor will ask for exactly those two things. Close the quarter by setting a recurring review. Microsoft documents access reviews in Entra ID as a way to review group memberships, application access and role assignments regularly so that only the right people keep access. Add the new boundary to that cycle, and add a pipeline check so the removed access is not granted again. The check keeps a quarter's work from being undone the following week. ## What to measure at day 90 Report four numbers: the share of accounts with logging verified, the count of unused roles and keys removed, the number of principals whose outcome changed under the enforced boundary, and the number of exceptions open past their expiry date. A fifth, if you have a reachability model, is the count of paths removed. These are plain counts that any team can produce, and they make a quarter's progress readable by someone outside the team. If the quarter stalls, [why least privilege stalls](/why-least-privilege-stalls/) lists the four usual stopping points. The next quarter repeats month two and three with the next class of access. The [routes page](/routes/) lists ways to get help with the work. ## Related pages - [Zero trust and least privilege: what NIST SP 800-207 says](/zero-trust-least-privilege/) - [Least privilege examples in AWS, Azure and Google Cloud](/least-privilege-examples/) - [IAM least privilege in AWS, Azure and Google Cloud](/iam-least-privilege/) ## Sources - Google Cloud, role recommendations overview: https://cloud.google.com/policy-intelligence/docs/role-recommendations-overview (read 16 Nov 2026) - AWS, generate policies based on access activity: https://docs.aws.amazon.com/IAM/latest/UserGuide/access-analyzer-policy-generation.html (read 16 Nov 2026) - NIST SP 800-53 Rev. 5 catalog (OSCAL JSON), control AC-6: https://raw.githubusercontent.com/usnistgov/oscal-content/main/nist.gov/SP800-53/rev5/json/NIST_SP-800-53_rev5_catalog.json (read 16 Nov 2026) - Microsoft, access reviews in Microsoft Entra ID: https://learn.microsoft.com/en-us/entra/id-governance/access-reviews-overview (read 16 Nov 2026) - AWS, testing IAM policies with the policy simulator: https://docs.aws.amazon.com/IAM/latest/UserGuide/access_policies_testing-policies.html (read 16 Nov 2026) - Google Cloud, Policy Simulator for allow policies: https://cloud.google.com/policy-intelligence/docs/iam-simulator-overview (read 16 Nov 2026) - AWS, IAM Access Analyzer findings: https://docs.aws.amazon.com/IAM/latest/UserGuide/access-analyzer-findings.html (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/first-90-days/ --- # Zero trust treats least privilege as a per-request decision, not a one-time grant Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. NIST SP 800-207 puts least privilege inside zero trust: its definition calls for accurate, least privilege per-request access decisions, and its third tenet grants access with the least privileges needed to complete the task. Least privilege is the access rule, and zero trust is the architecture that enforces it on each request. Read 16 Nov 2026. ## NIST defines zero trust in terms of least privilege NIST SP 800-207, Zero Trust Architecture, was published in August 2020. Its operative definition reads, verbatim: "Zero trust (ZT) provides a collection of concepts and ideas designed to minimize uncertainty in enforcing accurate, least privilege per-request access decisions in information systems and services in the face of a network viewed as compromised." The text is on the [NIST publication page](https://csrc.nist.gov/pubs/sp/800/207/final) and in the [full PDF](https://nvlpubs.nist.gov/nistpubs/SpecialPublications/NIST.SP.800-207.pdf), both read on 16 Nov 2026. The same section says the initial focus should be on restricting resources to those with a need to access and granting only the minimum privileges, with the examples read, write and delete, needed to perform the mission. It adds that access rules are made as granular as possible to enforce least privileges needed to perform the action in the request. ## Three of the seven tenets carry the rule Tenet 3 reads: "Access to individual enterprise resources is granted on a per-session basis." It continues that access should also be granted with the least privileges needed to complete the task, and that authentication and authorization to one resource will not automatically grant access to a different resource. Tenet 4 says access to resources is determined by dynamic policy, and states: "Least privilege principles are applied to restrict both visibility and accessibility." Tenet 6 says all resource authentication and authorization are dynamic and strictly enforced before access is allowed. Together they turn least privilege from a state of the account into a decision made at each request. ## The two ideas are related and not the same Least privilege is a rule about how much access a subject holds. Zero trust is an architecture that decides, for each request, whether the subject holds the access, using the subject's identity, the asset's state and other attributes. A zero trust deployment with wide grants still makes wide decisions, and a narrow grant behind a flat network is still exposed to the network. SP 800-207 also says that a collection of privileges given to multiple subjects could be thought of as a role, but that privileges should be assigned to a subject on an individual basis and not simply because they may fit into a particular role in the organization. That is stricter than most cloud role design, and it explains why the document keeps pushing toward attributes and conditions, covered on [least privilege access control](/least-privilege-access-control/). ## The migration advice applies to cloud access clean-up SP 800-207 notes that users with special privileges, such as developers or system administrators, require additional scrutiny when being assigned attributes or roles, and that in many legacy architectures such accounts hold broad permissions. It also says a new zero trust workflow could be operated in "reporting-only mode" for some time to make sure the policies are effective and workable. Reporting-only mode is the audit mode that cloud boundaries offer, and the scrutiny of administrator accounts is the first step in the [first 90 days](/first-90-days/). The NIST AC-6 requirement behind both is on [the principle of least privilege](/principle-of-least-privilege/). A practical reading for a cloud team: write the least privilege rule once, as policy and boundaries, and let the platform evaluate it on every request instead of relying on a review that runs quarterly. The review stays, since AC-6(7) asks for it, but it checks the rules rather than approving each call. That division of work is why the zero trust documents and the cloud IAM documents describe the same controls from two directions. ## Cloud providers give you the enforcement points The cloud mechanisms in the table are where the per-request decision is made. This site does not review zero trust products. It covers the access rules the clouds enforce. | Zero trust element in SP 800-207 | Cloud mechanism that enforces it | | --- | --- | | Least privileges for the task | Narrow roles and scoped policies | | Dynamic policy using attributes | ABAC tags, IAM Conditions, role assignment conditions | | Per-session access | Temporary credentials, time-bound assignments | | Reporting-only mode | Audit or dry-run modes for boundaries and simulators | ## Related pages - [Least privilege examples in AWS, Azure and Google Cloud](/least-privilege-examples/) - [IAM least privilege in AWS, Azure and Google Cloud](/iam-least-privilege/) - [Why least privilege projects stall in the cloud](/why-least-privilege-stalls/) ## Sources - NIST SP 800-207, Zero Trust Architecture (PDF): https://nvlpubs.nist.gov/nistpubs/SpecialPublications/NIST.SP.800-207.pdf (read 16 Nov 2026) - NIST CSRC, SP 800-207 publication page: https://csrc.nist.gov/pubs/sp/800/207/final (read 16 Nov 2026) - AWS, security best practices in IAM: https://docs.aws.amazon.com/IAM/latest/UserGuide/best-practices.html (read 16 Nov 2026) - Google Cloud, overview of IAM Conditions: https://cloud.google.com/iam/docs/conditions-overview (read 16 Nov 2026) - Microsoft, what is Privileged Identity Management: https://learn.microsoft.com/en-us/entra/id-governance/privileged-identity-management/pim-configure (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/zero-trust-least-privilege/ --- # Ten least privilege examples, each shown as a before and an after permission Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. Ten concrete cloud examples of least privilege, four on AWS, three on Azure and three on Google Cloud, each with the permission before, the permission after and the provider documentation that supports it. The examples are illustrative configurations, and the cited pages were read on 16 Nov 2026. ## The ten examples at a glance The before column is a common starting grant. The after column is the narrower form the cited page supports. Names such as example-bucket are placeholders. Check each change against your own usage first, with the steps in the [first 90 days](/first-90-days/). | # | Cloud | Before | After | Documentation | | --- | --- | --- | --- | --- | | 1 | AWS | Action s3:* on Resource * | s3:GetObject on one bucket prefix | [IAM best practices](https://docs.aws.amazon.com/IAM/latest/UserGuide/best-practices.html) | | 2 | AWS | AdministratorAccess on a developer role | A job function managed policy such as Developer power user | [Job function policies](https://docs.aws.amazon.com/IAM/latest/UserGuide/access_policies_job-functions.html) | | 3 | AWS | IAM user with a long-term access key in a pipeline | IAM role with temporary credentials | [IAM best practices](https://docs.aws.amazon.com/IAM/latest/UserGuide/best-practices.html) | | 4 | AWS | Lambda execution role with broad managed policy | Policy limited to the function's own actions and resources | [Lambda execution role](https://docs.aws.amazon.com/lambda/latest/dg/lambda-intro-execution-role.html) | | 5 | Azure | Owner at subscription scope | A narrower role at resource group scope | [Azure RBAC best practices](https://learn.microsoft.com/en-us/azure/role-based-access-control/best-practices) | | 6 | Azure | Permanent privileged assignment | Time-bound eligible assignment in PIM | [Privileged Identity Management](https://learn.microsoft.com/en-us/entra/id-governance/privileged-identity-management/pim-configure) | | 7 | Azure | Blob read granted on the whole storage account | Read granted with a condition on a blob tag | [Azure ABAC](https://learn.microsoft.com/en-us/azure/role-based-access-control/conditions-overview) | | 8 | Google Cloud | roles/editor on a project | Predefined roles for the services used | [Using IAM securely](https://cloud.google.com/iam/docs/using-iam-securely) | | 9 | Google Cloud | Default service account with Editor on the project | A dedicated service account per application | [Service account practices](https://cloud.google.com/iam/docs/best-practices-service-accounts) | | 10 | Google Cloud | Service account key stored outside Google Cloud | Workload identity federation | [Workload identity federation](https://cloud.google.com/iam/docs/workload-identity-federation) | ## AWS examples narrow actions, resources and credentials Example 1 follows AWS's advice to define the actions allowed on specific resources under specific conditions. A policy that allows s3:* on every resource lets the identity read, write and delete in any bucket. Replacing it with s3:GetObject on arn:aws:s3:::example-bucket/reports/* limits the identity to reading one set of objects. Example 2 swaps a full administrator policy for a job function policy, which AWS says aligns to common roles and carries filters that limit access. Example 3 removes a long-term key: AWS says workloads outside AWS can still use temporary credentials, which removes a secret that can leak. Example 4 follows the Lambda guidance to adjust the execution role before production, since permissions granted beyond what is required during development are common. ## Azure examples narrow scope, time and data Example 5 applies Microsoft's advice that, if a privileged administrator role must be assigned, a narrow scope such as a resource group or resource is better than a management group or subscription. Microsoft also suggests no more than three subscription owners. Example 6 replaces a standing assignment with one that is eligible, so the person activates it for a time-bound period and the access ends on its own. Example 7 uses a role assignment condition. Microsoft's own illustration is read access to blobs only if the blobs are tagged with a given project. It filters the permissions the role grants, so the assignment can stay at a wider scope while the effect is narrow. See [least privilege access control](/least-privilege-access-control/) for how conditions fit with roles. ## Google Cloud examples remove broad roles and keys Example 8 follows Google's guidance that basic roles include thousands of permissions across all services, and that role recommendations plus Policy Simulator can be used to decide which roles to grant instead and to confirm the change will not affect access. Example 9 addresses a default: Google says certain services create a default service account with the Editor role on the project, and recommends dedicated service accounts for each application. Example 10 removes the key altogether. Google says to avoid service account keys whenever possible, and workload identity federation lets an external workload use short-lived credentials instead. Each of the ten changes narrows one of the four properties on [least privilege access](/least-privilege-access/): action, resource, condition or time. The provider-by-provider view is on [IAM least privilege](/iam-least-privilege/). Two cautions apply to all ten. First, a narrower permission can break a job that runs rarely, so check the usage window before each change, as described on [privilege creep](/privilege-creep/). Second, the after column is a target, not a template: replace the placeholder bucket, role and project names with your own, and test the result in a non-production account before it reaches production. ## Related pages - [IAM least privilege in AWS, Azure and Google Cloud](/iam-least-privilege/) - [Why least privilege projects stall in the cloud](/why-least-privilege-stalls/) - [What a boundary is in cloud access control](/what-a-boundary-is/) ## Sources - AWS, security best practices in IAM: https://docs.aws.amazon.com/IAM/latest/UserGuide/best-practices.html (read 16 Nov 2026) - AWS, managed policies for job functions: https://docs.aws.amazon.com/IAM/latest/UserGuide/access_policies_job-functions.html (read 16 Nov 2026) - AWS Lambda, execution role: https://docs.aws.amazon.com/lambda/latest/dg/lambda-intro-execution-role.html (read 16 Nov 2026) - Microsoft, best practices for Azure RBAC: https://learn.microsoft.com/en-us/azure/role-based-access-control/best-practices (read 16 Nov 2026) - Microsoft, what is Privileged Identity Management: https://learn.microsoft.com/en-us/entra/id-governance/privileged-identity-management/pim-configure (read 16 Nov 2026) - Microsoft, what is Azure attribute-based access control: https://learn.microsoft.com/en-us/azure/role-based-access-control/conditions-overview (read 16 Nov 2026) - Google Cloud, using IAM securely: https://cloud.google.com/iam/docs/using-iam-securely (read 16 Nov 2026) - Google Cloud, best practices for service accounts: https://cloud.google.com/iam/docs/best-practices-service-accounts (read 16 Nov 2026) - Google Cloud, workload identity federation: https://cloud.google.com/iam/docs/workload-identity-federation (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/least-privilege-examples/ --- # IAM least privilege works the same in three clouds and is configured differently in each Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. IAM least privilege is the practice of limiting what identity and access management lets each identity do. AWS, Azure and Google Cloud each document a narrow role, a scope, a usage report and a ceiling, under different names. This page lines the four up, with documentation read on 16 Nov 2026. ## The same four controls exist in each cloud Each provider offers a way to pick a narrow role, to limit where it applies, to see what was used, and to cap what grants can allow. The names differ, and so do the traps, so a control that works in one cloud cannot be assumed to behave the same in another. | Control | AWS | Azure | Google Cloud | | --- | --- | --- | --- | | Narrow role | Customer managed policy or job function policy | Built-in or custom Azure role | Predefined or custom role | | Scope | Resource ARNs and conditions in the policy | Management group, subscription, resource group, resource | Organization, folder, project, resource | | Usage report | Last accessed information, Access Analyzer unused access | Defender for Cloud permissions management | Role recommendations | | Ceiling | Permissions boundary, service control policy | Deny assignment | Deny policy, Principal Access Boundary | ## AWS narrows by policy and reports by last use AWS puts the limits in the policy: actions on specific resources under specific conditions. Its job function managed policies are a starting point, and Access Analyzer can generate a policy from CloudTrail activity. The last accessed information shows service and management action use with a tracking period of at least 400 days for service information, and Access Analyzer unused access findings list unused roles, access keys, passwords and permissions. AWS says to regularly review and remove unused users, roles, permissions, policies and credentials. For the ceiling, permissions boundaries limit one user or role, and service control policies limit the accounts in an organization. Because the console shows one account at a time, run the reports at the organization level where the provider allows it, so that a forgotten account is not the one that holds the oldest role. The identity types are covered in [AWS human identities](/guides/aws-human-identities/) and [AWS service identities](/guides/aws-service-identities/). ## Azure narrows by role and scope and reports through Defender for Cloud Azure RBAC grants a role at a scope. Microsoft's guidance is to limit subscription owners to a maximum of 3, to avoid broader roles at broader scopes, and to use a narrow scope when a privileged administrator role must be assigned. Privileged Identity Management adds time-bound access, and access reviews in Entra ID repeat the check on a schedule. Microsoft says Defender for Cloud analyzes identity configurations and usage patterns to identify excessive, unused or misconfigured permissions for users, service principals, managed identities and service accounts. The ceiling is the deny assignment, which blocks actions even when a role assignment grants them. See [Azure human identities](/guides/azure-human-identities/) and [Azure workload identities](/guides/azure-workload-identities/). ## Google Cloud narrows by role and reports through recommendations Google separates basic roles, which hold thousands of permissions, from predefined roles and custom roles. Role recommendations compare granted permissions with those used in the last 90 days and suggest removing or replacing a role, and Policy Simulator shows the effect of a change before it is made. Google recommends dedicated service accounts for each application and warns about default service accounts that carry the Editor role. Its ceilings are deny policies, which apply regardless of granted roles, and Principal Access Boundary policies, which define the resources a set of principals can access. See [Google Cloud service identities](/guides/google-cloud-service-identities/) and [Google Cloud third-party identities](/guides/google-cloud-third-party-identities/). ## Three limits are shared by every provider report Read the reports with the same care in every cloud. All three reports are usage-based, so a role used less often than the window looks unused. None of the three applies a removal on its own: you apply the change, and you answer for its effect. And none sees access that is granted through a path outside IAM, such as a key stored in a secret manager. Treat the reports as a list of candidates, then test each change. Record the date each report was run, the window it covered and the person who accepted or dismissed each item, because the same report run next quarter will show different results as workloads change. The mechanisms behind the ceilings are on [least privilege access control](/least-privilege-access-control/), and ten before and after settings are in [least privilege examples](/least-privilege-examples/). ## Related pages - [Why least privilege projects stall in the cloud](/why-least-privilege-stalls/) - [What a boundary is in cloud access control](/what-a-boundary-is/) - [Principle of least privilege: definition and NIST AC-6](/principle-of-least-privilege/) ## Sources - AWS, security best practices in IAM: https://docs.aws.amazon.com/IAM/latest/UserGuide/best-practices.html (read 16 Nov 2026) - AWS, managed policies for job functions: https://docs.aws.amazon.com/IAM/latest/UserGuide/access_policies_job-functions.html (read 16 Nov 2026) - AWS, refine permissions using last accessed information: https://docs.aws.amazon.com/IAM/latest/UserGuide/access_policies_last-accessed.html (read 16 Nov 2026) - AWS, IAM Access Analyzer findings: https://docs.aws.amazon.com/IAM/latest/UserGuide/access-analyzer-findings.html (read 16 Nov 2026) - AWS, permissions boundaries for IAM entities: https://docs.aws.amazon.com/IAM/latest/UserGuide/access_policies_boundaries.html (read 16 Nov 2026) - AWS, service control policies: https://docs.aws.amazon.com/organizations/latest/userguide/orgs_manage_policies_scps.html (read 16 Nov 2026) - Microsoft, best practices for Azure RBAC: https://learn.microsoft.com/en-us/azure/role-based-access-control/best-practices (read 16 Nov 2026) - Microsoft, what is Privileged Identity Management: https://learn.microsoft.com/en-us/entra/id-governance/privileged-identity-management/pim-configure (read 16 Nov 2026) - Microsoft, permissions management in Defender for Cloud: https://learn.microsoft.com/en-us/azure/defender-for-cloud/permissions-management (read 16 Nov 2026) - Microsoft, list Azure deny assignments: https://learn.microsoft.com/en-us/azure/role-based-access-control/deny-assignments (read 16 Nov 2026) - Google Cloud, using IAM securely: https://cloud.google.com/iam/docs/using-iam-securely (read 16 Nov 2026) - Google Cloud, role recommendations overview: https://cloud.google.com/policy-intelligence/docs/role-recommendations-overview (read 16 Nov 2026) - Google Cloud, best practices for service accounts: https://cloud.google.com/iam/docs/best-practices-service-accounts (read 16 Nov 2026) - Google Cloud, deny policies: https://cloud.google.com/iam/docs/deny-overview (read 16 Nov 2026) - Google Cloud, Principal Access Boundary policies: https://cloud.google.com/iam/docs/principal-access-boundary-policies (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/iam-least-privilege/ --- # Least privilege projects stall because removing access carries more risk than granting it Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. Cloud least privilege projects stall at four points: nobody can show a removal is safe, usage data covers too short a window, clean-up happens one permission at a time, and the report ends in a ticket. Provider documentation read on 16 Nov 2026 describes a way past each of the four. ## Granting is cheap and removal has an owner who fears the outage AWS documents the pattern in its own IAM guidance: you might start with broad permissions while you explore what a workload needs, and as the use case matures you work to reduce them. The first half of that sentence happens on every project. The second half needs someone to decide that a removal is safe, and that person carries the outage if the decision is wrong. That asymmetry is the first reason projects stall. A broad grant costs nothing to approve and a failed deploy is visible within minutes. A removal that breaks a quarterly job is visible three months later, with the engineer who approved it as the named cause. Teams respond by freezing the clean-up queue. The way past it is evidence per change: a record of what the change would have blocked, before it ships. ## The usage window decides what looks unused Usage-based tools compare what is granted with what was used in a window. Google documents that the IAM recommender compares a principal's permissions with those used in the last 90 days, and that by default it does not generate recommendations until it has 90 days of data. For project-level recommendations the minimum can be set to 30 or 60 days, and Google says accuracy might be affected. AWS lets policy generation read up to 90 days of CloudTrail events. A role used once a year for failover, or a job that runs at quarter close, can sit outside any 90-day window. Removing its permissions because they look unused is the mistake that makes the next removal harder to approve. Treat the window as a stated limit of the data, and list the identities whose duties run on longer cycles before any removal. [Privilege creep](/privilege-creep/) lists the other windows. ## Permission-by-permission clean-up does not scale Google's role recommendations each suggest that you remove or replace a role for a principal. That is the right unit for a review and the wrong unit for an estate of hundreds of accounts. If every finding is a ticket, the queue grows at the rate that engineers create roles, which is faster than anyone closes tickets. A boundary is the alternative unit: one reviewed limit that acts on every identity beneath it, covered on [what a boundary is](/what-a-boundary-is/). The [Access Debt Calculator](/) shows the size of the difference for your own counts, and its arithmetic is on the [methodology page](/methodology/). ## A finding ends the tool's work and starts the team's AWS Access Analyzer recommends a refined policy for unused permissions and the user attaches it. Google states that the IAM recommender does not apply recommendations automatically, and that you apply or dismiss each one. Both are sound designs for a provider tool. A recommendation service recommends, and the remaining work is staffing: someone must read each result, test it and ship it. Whoever owns the last step also owns the risk of the change, so the tool's own work is not the whole project. Name the owner of the last step before the first finding arrives, and put it in the written [policy](/least-privilege-policy-template/). ## New access arrives faster than old access leaves Even a clean estate decays. Workloads and identities are created and removed by pipelines at any hour, so a clean-up is a snapshot, and the next deploy can re-grant what was just removed. The durable answer is a check before production: a policy test in the pipeline, with exceptions handled in the team's own tools. AWS documents custom policy checks in IAM Access Analyzer, which check a policy for new access against your security standards and can be run from the CLI or API as well as the console. A project with no check before production is cleaning a floor while the tap runs. The [first 90 days](/first-90-days/) ends with such a check for that reason. Count the stall by its stage. If findings exist and no change has been shipped, the stall is the first or second point and needs an owner and a longer window. If changes ship and then return, the stall is the fifth point and needs a check in the pipeline. If every change takes a week of meetings, the unit is wrong and a boundary is the better candidate. Each diagnosis points to a different next step, so naming the stage saves a quarter spent on the wrong one. ## Related pages - [What a boundary is in cloud access control](/what-a-boundary-is/) - [Principle of least privilege: definition and NIST AC-6](/principle-of-least-privilege/) - [Privilege creep in the cloud: causes and how to measure it](/privilege-creep/) ## Sources - AWS, security best practices in IAM: https://docs.aws.amazon.com/IAM/latest/UserGuide/best-practices.html (read 16 Nov 2026) - Google Cloud, role recommendations overview: https://cloud.google.com/policy-intelligence/docs/role-recommendations-overview (read 16 Nov 2026) - AWS, generate policies based on access activity: https://docs.aws.amazon.com/IAM/latest/UserGuide/access-analyzer-policy-generation.html (read 16 Nov 2026) - AWS, IAM Access Analyzer findings: https://docs.aws.amazon.com/IAM/latest/UserGuide/access-analyzer-findings.html (read 16 Nov 2026) - AWS, IAM Access Analyzer custom policy checks: https://docs.aws.amazon.com/IAM/latest/UserGuide/access-analyzer-custom-policy-checks.html (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/why-least-privilege-stalls/ --- # A boundary is a limit set above the grants, so one change can remove a class of access Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. A boundary is a rule that caps what the grants below it can do, whatever those grants say. AWS, Google Cloud and Azure each document one: a ceiling on a role, a limit on the resources a principal may reach, or a deny that overrides assignments. One boundary can remove many paths at once. ## A grant gives access and a boundary limits it A grant says an identity may do something. A boundary says what no grant beneath it can allow. The two are evaluated together, and the effective result is what both permit. That is why a boundary can remove access without touching the grants: the broad role stays in place and the boundary makes parts of it inert. The same shape appears in every cloud, though the names and evaluation rules differ. This page uses the word in the general sense and names one example per cloud to show the shape. A reader should not expect the three examples below to behave alike, since each cloud evaluates its own policy types in its own order. The mechanisms side by side are on [least privilege access control](/least-privilege-access-control/). ## Each cloud documents its own version AWS documents a permissions boundary as an advanced feature that uses a managed policy to set the maximum permissions an identity-based policy can grant to an IAM user or role. The entity can perform only the actions allowed by both its identity-based policies and its boundary. It is a ceiling on one identity and grants nothing itself. Google Cloud documents Principal Access Boundary policies as policies that define the resources principals can access. They attach to principal sets, in contrast to allow and deny policies, which attach to resources. Google gives the example of preventing principals from accessing resources in other organizations, and notes that by default principals are eligible to access any Google Cloud resource, so the boundary narrows an otherwise open default. Microsoft documents deny assignments in Azure RBAC as attaching a set of deny actions to a user, group or service principal at a scope, blocking those actions even if a role assignment grants them. The page also covers listing them, which is the first step in learning which denies already exist in your tenant. ## A boundary removes a class and a clean-up removes an item A permission-by-permission clean-up deletes one grant per change and has to be repeated for every identity. A boundary applied at an organization, folder or account level acts on every identity beneath it at once. If the boundary denies a service nobody uses, a region nobody runs in, or access from outside the organization, thousands of grants become irrelevant in one change. The comparison is between ten thousand edits and one reviewed policy. The [Access Debt Calculator](/) puts numbers on that comparison: it sizes a permission-by-permission plan and a boundary plan from the same counts, and shows the arithmetic on the [methodology page](/methodology/). [Why least privilege stalls](/why-least-privilege-stalls/) explains why the first plan rarely finishes. ## A boundary needs a simulation more than a grant does Because a boundary acts on many identities, its mistakes are wide. A deny that is slightly too broad blocks a deployment role in forty accounts at the same moment. The same reach that makes a boundary efficient makes an untested one dangerous. Release should be gradual for the same reason, starting with the accounts that have the fewest dependencies, and the list of affected principals needs an owner for each. The practical rule is to replay recent access under the proposed boundary, list every principal that would change outcome, and release it in audit mode first. Google's Policy Intelligence includes simulators for deny policies and Principal Access Boundary policies, and AWS provides a policy simulator for testing IAM policies. ## A boundary should be written down in version control Whichever cloud enforces it, the audit question is the same: where is the boundary written down, who can change it, and how would you know that someone did. A boundary stored only in a console is a boundary with no history. Keeping it as code in version control is the simplest answer, since a pull request records the author, the reviewer and the date. The written [policy template](/least-privilege-policy-template/) puts these three answers in section 5. ## Related pages - [Principle of least privilege: definition and NIST AC-6](/principle-of-least-privilege/) - [Privilege creep in the cloud: causes and how to measure it](/privilege-creep/) - [Least privilege access: how access is granted and limited](/least-privilege-access/) ## Sources - AWS, permissions boundaries for IAM entities: https://docs.aws.amazon.com/IAM/latest/UserGuide/access_policies_boundaries.html (read 16 Nov 2026) - Google Cloud, Principal Access Boundary policies: https://cloud.google.com/iam/docs/principal-access-boundary-policies (read 16 Nov 2026) - Microsoft, list Azure deny assignments: https://learn.microsoft.com/en-us/azure/role-based-access-control/deny-assignments (read 16 Nov 2026) - AWS, testing IAM policies with the policy simulator: https://docs.aws.amazon.com/IAM/latest/UserGuide/access_policies_testing-policies.html (read 16 Nov 2026) - Google Cloud, Policy Simulator for allow policies: https://cloud.google.com/policy-intelligence/docs/iam-simulator-overview (read 16 Nov 2026) - AWS, service control policies: https://docs.aws.amazon.com/organizations/latest/userguide/orgs_manage_policies_scps.html (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/what-a-boundary-is/ --- # Least privilege for 20 common cloud services Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. Services describes the help available to a team running a least privilege program in AWS, Azure or Google Cloud, what each kind of help covers and what the team still has to provide. The page lists what it covers by task, states the date it was written, and ranks no provider or product. - [Least privilege for Amazon S3](/services/aws-s3/): Prefix-scoped grants, Block Public Access and what S3 reports about use. - [Least privilege for AWS Lambda](/services/aws-lambda/): Execution role, resource-based policy and Access Analyzer trimming. - [Least privilege for Amazon EC2](/services/aws-ec2/): Instance roles, Describe actions and the metadata service. - [Least privilege for Amazon EKS](/services/aws-eks/): Access entries, access policies and Pod Identity. - [Least privilege for Amazon RDS](/services/aws-rds/): Two kinds of RDS permissions and the ARN that scopes a login. - [Least privilege for Amazon DynamoDB](/services/aws-dynamodb/): Table ARNs, index ARNs and fine-grained condition keys. - [Least privilege for AWS KMS](/services/aws-kms/): Key policies, administrators versus users and encryption context. - [Least privilege for AWS Secrets Manager](/services/aws-secrets-manager/): Per-secret grants, the KMS key and the administrator pair. - [Least privilege for Azure Storage](/services/azure-storage/): Data roles, container scope, no account keys. - [Least privilege for Azure Key Vault](/services/azure-key-vault/): Two planes, one vault per app and environment. - [Least privilege for Azure Kubernetes Service](/services/azure-aks/): Four built-in roles, namespace scope. - [Least privilege for Azure Functions](/services/azure-functions/): Three roles on the app, identity for everything else. - [Least privilege for Azure SQL Database](/services/azure-sql-database/): Azure RBAC for the server, database roles for the data. - [Least privilege for Cloud Storage](/services/gcp-cloud-storage/): Bucket-level roles, no basic roles. - [Least privilege for BigQuery](/services/gcp-bigquery/): Run jobs in one place, read data in another. - [Least privilege for Google Kubernetes Engine](/services/gcp-gke/): IAM for the project, RBAC for the cluster. - [Least privilege for Cloud Run](/services/gcp-cloud-run/): Deploy rights, invoke rights and the identity are three grants. - [Least privilege for Cloud Functions](/services/gcp-cloud-functions/): Administer, invoke and deploy are separate rights. - [Least privilege for Cloud KMS](/services/gcp-cloud-kms/): Grant on the key, not the project. - [Least privilege for Pub/Sub](/services/gcp-pubsub/): Per-topic and per-subscription roles. ## Related pages - [Routes to least privilege: tools and platforms](/routes/) - [The least privilege data note: every edition](/data-note/) - [Guides](/guides/) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/services/ --- # Least privilege for Amazon S3 is a prefix-scoped grant with ACLs off and public access blocked Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. AWS documentation read on 16 Nov 2026 says new S3 buckets disable ACLs by default, so access rests on identity policies, bucket policies, VPC endpoint policies, SCPs and RCPs. This guide shows a prefix-scoped grant, the wildcards AWS tells you to hunt for, and the checks for what is used. ## S3 access rests on policies once ACLs are disabled AWS documentation says S3 Object Ownership defaults to the bucket owner enforced setting, which disables all access control lists (ACLs). When ACLs are off, access control is based on policies: IAM user policies, S3 bucket policies, VPC endpoint policies, AWS Organizations service control policies (SCPs) and resource control policies (RCPs). The same page recommends granting only the permissions required for a task and says least privilege reduces the impact of errors or malicious intent. It also recommends IAM roles with temporary credentials over long-term credentials stored in an application or on an EC2 instance. The identities that call this service (people, services, workloads, third parties) are covered in the [AWS service identities guide](/guides/aws-service-identities/) and the [workload identities guide](/guides/aws-workload-identities/). ## Wildcard principals, wildcard actions and ACL grants are the patterns to remove AWS documentation tells you to look for four patterns, and gives its reason for each. - A bucket policy with "Principal": "*", which effectively means anyone. - A bucket policy with a wildcard action "*", which effectively allows every action on the bucket. - Bucket ACLs that give read, write or full access to "Everyone" or "Any authenticated AWS user". - SSE-C encryption left enabled when no workload needs it: AWS says it makes access impractical to share with other users, roles or services, and since April 2026 it is disabled by default on new general purpose buckets. ## A home-folder grant limits one user to one prefix AWS publishes an example bucket policy that confines a user named JohnDoe to the home/JohnDoe/ prefix. The example grants s3:* on the prefix; a task that only reads and writes objects can name those object actions instead. | Statement | Action | Resource | Condition | | --- | --- | --- | --- | | AllowRootAndHomeListingOfCompanyBucket | s3:ListBucket | arn:aws:s3:::amzn-s3-demo-bucket | s3:prefix is "", "home/" or "home/JohnDoe"; s3:delimiter is "/" | | AllowListingOfUserFolder | s3:ListBucket | arn:aws:s3:::amzn-s3-demo-bucket | s3:prefix matches home/JohnDoe/* | | AllowAllS3ActionsInUserFolder | s3:* (in the AWS example) | arn:aws:s3:::amzn-s3-demo-bucket/home/JohnDoe/* | none | ## Bucket-level controls narrow access beyond identity policies - S3 Block Public Access applies centralized limits regardless of how a resource was created, and AWS says an organization-level policy can apply it to all member accounts or selected OUs. - A Deny statement with the aws:PrincipalOrgID condition key keeps access inside an AWS Organization; AWS calls this stronger than an Allow with the same key. - A Deny with aws:SecureTransport set to "false" rejects requests sent over HTTP. - The s3:x-amz-acl condition key requires uploads to carry the bucket-owner-full-control ACL. ## Check use with Access Analyzer, CloudTrail data events and the 403 message IAM Access Analyzer for S3 reports buckets open to the internet or to other AWS accounts, and AWS says the external access summary refreshes once every 24 hours. AWS Config managed rules s3-bucket-public-read-prohibited and s3-bucket-public-write-prohibited give ongoing detection. Object-level activity such as GetObject, ListObjects and PutObject appears in CloudTrail data events. AWS documentation on IAM last accessed information says action-level data covers management actions and not data plane actions, and the policy generation page says Access Analyzer does not identify S3 data events in generated policies. Use CloudTrail data events for who read which prefix. ## Related pages - [Least privilege for AWS Lambda](/services/aws-lambda/) - [Least privilege for Amazon EC2](/services/aws-ec2/) - [Least privilege for Amazon EKS](/services/aws-eks/) - [All of them](/services/) ## Sources - Security best practices for Amazon S3: https://docs.aws.amazon.com/AmazonS3/latest/userguide/security-best-practices.html (read 16 Nov 2026) - Bucket policy examples: https://docs.aws.amazon.com/AmazonS3/latest/userguide/example-bucket-policies.html (read 16 Nov 2026) - IAM Access Analyzer for S3: https://docs.aws.amazon.com/AmazonS3/latest/userguide/access-analyzer.html (read 16 Nov 2026) - IAM last accessed information: https://docs.aws.amazon.com/IAM/latest/UserGuide/access_policies_last-accessed.html (read 16 Nov 2026) - IAM Access Analyzer policy generation: https://docs.aws.amazon.com/IAM/latest/UserGuide/access-analyzer-policy-generation.html (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/services/aws-s3/ --- # Least privilege for AWS Lambda starts with a narrow execution role and a scoped resource policy Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. AWS Lambda has two permission paths: the execution role a function uses and the resource-based policy that lets others invoke it. AWS documentation read on 16 Nov 2026 supports full JSON resource policies with condition keys. This guide covers both paths, an invoke grant for S3, and how to trim a role. ## Lambda has an execution role and a resource-based policy AWS documentation describes two categories of permissions: what a function needs to call other AWS resources, and what other entities need to reach the function. The first is defined in the execution role, which every function must have. At a minimum the role needs CloudWatch Logs access, which the AWSLambdaBasicExecutionRole managed policy provides. The second is set with identity-based policies or resource-based policies. AWS says that when a user calls a function Lambda considers both policy types, and that when an AWS service such as S3 calls a function Lambda considers only the resource-based policy. Attribute-based access control with tags is also supported. The identities that call this service (people, services, workloads, third parties) are covered in the [AWS service identities guide](/guides/aws-service-identities/) and the [workload identities guide](/guides/aws-workload-identities/). ## Development-time breadth and manual role assumption are the habits to avoid - Leaving the permissions from the development phase in place. AWS says you may grant more than required at first and should adjust the policy before publishing to production. - Starting from an AWS managed policy and never narrowing it. AWS recommends managed policies as a starting point for writing more restrictive ones; one example is AWSLambdaDynamoDBExecutionRole, which a function that reads a DynamoDB stream may use. - Calling sts:AssumeRole in function code to assume the execution role. AWS says to avoid it, because Lambda assumes the role automatically on invocation. - Using AddPermission statements when you need more control. AWS says individual permissions support only aws:SourceArn, aws:SourceAccount and aws:PrincipalOrgID and recommends full JSON policies. ## An invoke grant for S3 names the account and denies one bucket AWS publishes a resource-based policy that lets every S3 bucket in one account invoke a function except one bucket that is explicitly denied. Setting it needs lambda:PutResourcePolicy, lambda:AddPermission and lambda:RemovePermission. | Effect | Principal | Action | Condition | | --- | --- | --- | --- | | Allow | Service: s3.amazonaws.com | lambda:InvokeFunction on the function ARN | StringEquals aws:SourceAccount is the account ID | | Deny | Service: s3.amazonaws.com | lambda:InvokeFunction on the function and its versions | ArnLike aws:SourceArn is arn:aws:s3:::amzn-s3-demo-bucket | ## Full JSON policies add explicit denies and organization conditions AWS documentation says full JSON resource policies, up to 20 KB, can hold multiple principals, explicit Deny statements and the complete range of IAM global condition keys. It lists aws:PrincipalOrgID and aws:PrincipalOrgPaths as ways to restrict access to your organization without listing each account. AWS warns that put-resource-policy replaces any existing policy. Read the current one with get-resource-policy and pass its RevisionId so an older revision is not applied. ## Generate a trimmed role from CloudTrail activity AWS says IAM Access Analyzer reviews CloudTrail logs over a date range you choose and generates a policy template with only the permissions the function used. The policy generation page allows up to 90 days. You turn the template into a managed policy and attach it to the role. The same page says generated policies leave out action-level activity for data events. Check the output against what the function does at runtime before replacing the role policy. ## Related pages - [Least privilege for Amazon EC2](/services/aws-ec2/) - [Least privilege for Amazon EKS](/services/aws-eks/) - [Least privilege for Amazon RDS](/services/aws-rds/) - [All of them](/services/) ## Sources - Managing permissions in AWS Lambda: https://docs.aws.amazon.com/lambda/latest/dg/lambda-permissions.html (read 16 Nov 2026) - Lambda execution role: https://docs.aws.amazon.com/lambda/latest/dg/lambda-intro-execution-role.html (read 16 Nov 2026) - Resource-based policies in Lambda: https://docs.aws.amazon.com/lambda/latest/dg/access-control-resource-based.html (read 16 Nov 2026) - IAM Access Analyzer policy generation: https://docs.aws.amazon.com/IAM/latest/UserGuide/access-analyzer-policy-generation.html (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/services/aws-lambda/ --- # Least privilege for Amazon EC2 means a scoped instance role, split API actions and IMDSv2 Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. Amazon EC2 access has two sides: the IAM policies that let people call the EC2 API, and the instance role that applications use. AWS documentation read on 16 Nov 2026 supports narrowing Describe actions, one role per instance profile and policies generated from CloudTrail. This guide gives a read-only grant. ## EC2 permissions split into API access and instance roles AWS documentation says you control who can use EC2 resources with IAM policies and control access to instances with security groups. Applications on an instance get credentials from an IAM role attached through an instance profile, retrieved from the Instance Metadata Service (IMDS). An instance profile can hold only one IAM role, and an instance can have only one role attached, though many instances can share a role. AWS says to associate least privilege policies that restrict the role to the specific API calls the application requires. The identities that call this service (people, services, workloads, third parties) are covered in the [AWS service identities guide](/guides/aws-service-identities/) and the [workload identities guide](/guides/aws-workload-identities/). ## Describe wildcards and role swaps are the broad actions to watch - ec2:Describe* in a policy. AWS says Describe actions do not support resource-level permissions, so Resource must be "*"; replacing the wildcard with specific Describe actions is the way to shrink what a user can view. - Granting the launch wizard more than it needs. AWS lists ec2:RunInstances plus ec2:DescribeImages, ec2:DescribeInstanceTypes, ec2:DescribeVpcs, ec2:DescribeSubnets and a security group action as the minimum for a launch. - Letting users attach or replace roles without limits. AWS says resource-level permissions can control who may attach, replace or detach IAM roles for an instance. - Removing a role from an instance profile to revoke access. AWS says the change can take up to one hour to take effect and recommends replacing the instance profile. ## A read-only grant for instances, images and snapshots lists four actions AWS publishes this example for a user who only needs to view instances, AMIs and snapshots in the console. Dropping ec2:DescribeTags limits the view to private AMIs. | Effect | Action | Resource | Note from AWS | | --- | --- | --- | --- | | Allow | ec2:DescribeInstances | * | Describe actions need Resource "*" | | Allow | ec2:DescribeImages | * | Lists AMIs | | Allow | ec2:DescribeTags | * | Lets the console show public AMIs | | Allow | ec2:DescribeSnapshots | * | Lists snapshots | ## Instance metadata and the instance identity role limit what a host can hand out AWS documentation says that by default instances accept IMDSv1 and IMDSv2, and that you can require IMDSv2 so IMDSv1 calls fail. IMDSv2 uses session-oriented requests. Each instance also has an instance identity role with no trust policy, usable only by integrated services such as Systems Manager, GuardDuty Runtime Monitoring and EC2 Instance Connect. Because the credentials are present whether or not a service uses them, they are not a path for your own applications. ## Policy generation and CloudTrail show which calls an instance role made AWS says that before production you can generate an IAM policy from the activity of an IAM role. IAM Access Analyzer reviews your CloudTrail logs and creates a template containing the permissions used during the date range you specify. To find which API actions a console task needs, AWS suggests a logging service such as CloudTrail. If a policy blocks an action, the console shows an encoded message that the STS DecodeAuthorizationMessage action can decode. ## Related pages - [Least privilege for Amazon EKS](/services/aws-eks/) - [Least privilege for Amazon RDS](/services/aws-rds/) - [Least privilege for Amazon DynamoDB](/services/aws-dynamodb/) - [All of them](/services/) ## Sources - IAM roles for Amazon EC2: https://docs.aws.amazon.com/AWSEC2/latest/UserGuide/iam-roles-for-amazon-ec2.html (read 16 Nov 2026) - Example policies for the EC2 console: https://docs.aws.amazon.com/AWSEC2/latest/UserGuide/iam-policies-ec2-console.html (read 16 Nov 2026) - Instance metadata service: https://docs.aws.amazon.com/AWSEC2/latest/UserGuide/configuring-instance-metadata-service.html (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/services/aws-ec2/ --- # Least privilege for Amazon EKS means namespace-scoped access policies and one role per service account Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. Amazon EKS has two layers: AWS IAM decides who can call the EKS API, and Kubernetes permissions decide what they do inside the cluster. AWS documentation read on 16 Nov 2026 describes access entries scoped to namespaces and Pod Identity for pods. This guide shows a view-only namespace grant. ## EKS access entries map IAM principals to Kubernetes permissions AWS documentation says you can associate one or more access policies with access entries of type STANDARD. EKS access policies contain Kubernetes permissions, not IAM permissions. Access entries also let you modify or revoke the cluster-admin access given to the cluster creator. The four access policies AWS lists are AmazonEKSAdminPolicy, AmazonEKSClusterAdminPolicy, AmazonEKSEditPolicy and AmazonEKSViewPolicy. If a principal has several policies, it receives all permissions in all of them. Each policy can be associated with an access entry only once, and each association carries its own scope, so a principal can hold view access cluster-wide and admin access in one namespace. The identities that call this service (people, services, workloads, third parties) are covered in the [AWS service identities guide](/guides/aws-service-identities/) and the [workload identities guide](/guides/aws-workload-identities/). ## Cluster-wide scope and creator admin are the grants to avoid - A cluster-scoped association of AmazonEKSClusterAdminPolicy or AmazonEKSAdminPolicy when a namespace scope would do. AWS says an access policy can be scoped to all resources or to named namespaces, including wildcards such as dev-*. - Leaving the cluster creator with cluster-admin. AWS says access entries let you modify or revoke it. - Running pods with access to the node role. AWS says that if IMDS is not restricted, a pod can reach the node IAM role and may reach the roles of other pods. - Sharing one role across all workloads. AWS says Pod Identity lets you scope IAM permissions to a service account, so only pods using it get them. ## A view-only grant for two namespaces uses one CLI call AWS documents associating AmazonEKSViewPolicy with a role for named namespaces. The principal needs ListAccessEntries, DescribeAccessEntry, UpdateAccessEntry, ListAccessPolicies, AssociateAccessPolicy and DisassociateAccessPolicy to manage associations. | Item | Value from the AWS example | | --- | --- | | Access policy | arn:aws:eks::aws:cluster-access-policy/AmazonEKSViewPolicy | | Command | aws eks associate-access-policy | | Principal | arn:aws:iam::111122223333:role/my-role | | Access scope | type=namespace,namespaces=my-namespace1,my-namespace2 | | Widen later | Re-run with type=cluster; remove a namespace by re-running without it | ## Pod Identity and credential isolation narrow what a workload can obtain AWS says each EKS Pod Identity association maps one IAM role to one service account in a namespace, and a service account can be associated with one role from the cluster account. Identical associations can be made in several clusters without editing the role trust policy. Up to 5,000 associations per cluster are supported. AWS also says containers are not a security boundary: pods on the same node share a kernel. Pods with hostNetwork set to true always have IMDS access, even though SDKs use Pod Identity credentials when enabled. ## List associations and read CloudTrail to see what is granted and used Run aws eks list-associated-access-policies to see each policy and its scope for a principal. AWS warns that kubectl auth can-i --list does not show permissions granted by access policies, and that impersonating a user with --as forces Kubernetes RBAC so access policy permissions no longer apply. AWS lists access and event logging through CloudTrail as the Pod Identity audit source. Use IAM Access Analyzer policy generation on a pod role to compare granted and used actions, and review the access entries list after each team change. ## Related pages - [Least privilege for Amazon RDS](/services/aws-rds/) - [Least privilege for Amazon DynamoDB](/services/aws-dynamodb/) - [Least privilege for AWS KMS](/services/aws-kms/) - [All of them](/services/) ## Sources - Access entries: https://docs.aws.amazon.com/eks/latest/userguide/access-entries.html (read 16 Nov 2026) - Associate access policies with access entries: https://docs.aws.amazon.com/eks/latest/userguide/access-policies.html (read 16 Nov 2026) - EKS Pod Identity: https://docs.aws.amazon.com/eks/latest/userguide/pod-identities.html (read 16 Nov 2026) - IAM Access Analyzer policy generation: https://docs.aws.amazon.com/IAM/latest/UserGuide/access-analyzer-policy-generation.html (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/services/aws-eks/ --- # Least privilege for Amazon RDS pairs scoped rds: actions with rds-db:connect on named database users Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. Amazon RDS has control-plane permissions for managing instances and a separate rds-db:connect action for logging in with IAM database authentication. AWS documentation read on 16 Nov 2026 shows the ARN format that scopes a login to one database account. This guide covers that grant, tags and checks. ## RDS permissions cover management actions and database logins AWS documentation says RDS uses the rds: prefix for management API actions, such as rds:DescribeDBInstances, and the rds-db: prefix only for IAM database authentication. The rds-db:connect action is not valid in any other context. By default, permission sets and roles cannot create or modify RDS resources, and an administrator must attach policies naming the operations and resources. AWS recommends starting with AWS managed policies and moving to customer managed policies specific to the use case. The RDS examples page also covers policies to view your own permissions, to create, modify and delete resources, and to use condition keys and custom tags. The DbiResourceId in an rds-db ARN is unique to a Region and never changes, so a policy keeps working if the instance identifier is renamed. The identities that call this service (people, services, workloads, third parties) are covered in the [AWS service identities guide](/guides/aws-service-identities/) and the [workload identities guide](/guides/aws-workload-identities/). ## Account-wide dbuser wildcards and administrator access are the grants to narrow - The resource arn:aws:rds-db:us-east-2:111122223333:dbuser:*/*, which AWS shows as matching all DB instances and database accounts in one account and Region. - Administrator permissions in general. AWS says a user with administrator permissions can reach DB instances without explicit permissions in an IAM policy, and suggests a lesser-privileged role for administrators if you want to restrict that. - A console policy tighter than the console minimum. AWS says the console will not work as intended for principals whose policy is more restrictive than the minimum permissions. ## A database login for one account uses a single rds-db:connect ARN AWS publishes a policy with one statement. The ARN names the Region, the account, the DB instance resource ID and the database user. | Element | Value from the AWS example | | --- | --- | | Effect | Allow | | Action | rds-db:connect | | Resource | arn:aws:rds-db:us-east-2:111122223333:dbuser:db-ABCDEFGHIJKL01234/db_user | | ARN format | arn:aws:rds-db:region:account-id:dbuser:DbiResourceId/db-user-name | | Aurora and proxies | Use DbClusterResourceId for Aurora, or the proxy resource ID such as prx-ABCDEFGHIJKL01234 for RDS Proxy | ## Conditions, tags and cross-account roles narrow access further - Conditions: AWS says a policy condition can require SSL or allow an action only when used through a specific service such as CloudFormation. - Tags: AWS describes attribute-based access control with RDS tags in policies, and a policy that grants permission to tag resources at creation. - MFA: AWS says to add MFA conditions where IAM users or the root user are required. - Cross-account: AWS says to create the role with the connect policy in the DB account and let the other account assume it, because the user must be in the same account as the instance. ## Validate the policy and trim it from observed activity AWS says IAM Access Analyzer validates policies with more than 100 policy checks and gives actionable recommendations. Run it on the rds: and rds-db: statements before attaching them. IAM Access Analyzer policy generation reviews CloudTrail activity for a role and builds a template from the permissions used, for up to 90 days. That covers roles that call RDS management APIs. AWS notes the generator does not identify action-level activity for data events, so a history of database logins needs database-side logs. ## Related pages - [Least privilege for Amazon DynamoDB](/services/aws-dynamodb/) - [Least privilege for AWS KMS](/services/aws-kms/) - [Least privilege for AWS Secrets Manager](/services/aws-secrets-manager/) - [All of them](/services/) ## Sources - IAM policy for IAM database access: https://docs.aws.amazon.com/AmazonRDS/latest/UserGuide/UsingWithRDS.IAMDBAuth.IAMPolicy.html (read 16 Nov 2026) - Identity-based policy examples for RDS: https://docs.aws.amazon.com/AmazonRDS/latest/UserGuide/security_iam_id-based-policy-examples.html (read 16 Nov 2026) - How RDS works with IAM: https://docs.aws.amazon.com/AmazonRDS/latest/UserGuide/security_iam_service-with-iam.html (read 16 Nov 2026) - IAM Access Analyzer policy generation: https://docs.aws.amazon.com/IAM/latest/UserGuide/access-analyzer-policy-generation.html (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/services/aws-rds/ --- # Least privilege for Amazon DynamoDB is a table-scoped action list plus LeadingKeys and Attributes conditions Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. Amazon DynamoDB access is granted per table and per action, and AWS documentation read on 16 Nov 2026 adds condition keys for item and attribute level control. This guide shows a read-only table grant, the managed policies to replace, the dynamodb:LeadingKeys pattern and how to check use. ## DynamoDB permissions attach to table and index ARNs AWS documentation says users and roles have no permission to create or modify DynamoDB resources by default, and an administrator writes IAM policies to grant it. A statement names actions, a Resource ARN for a table in a Region and account, and optional conditions. AWS advises starting with AWS managed policies and then defining customer managed policies specific to the use case, and it lists more than 100 Access Analyzer policy checks as a validation step. The identities that call this service (people, services, workloads, third parties) are covered in the [AWS service identities guide](/guides/aws-service-identities/) and the [workload identities guide](/guides/aws-workload-identities/). ## The full-access managed policy grants more than most workloads need - AmazonDynamoDBFullAccess. AWS describes it as full access to DynamoDB resources through the console, and it has a read-only counterpart for viewing. A workload that touches one table does not need either; AWS shows a customer managed policy for that table instead. - A Resource of "*" on data actions. AWS example policies name a table ARN, and an index needs its own ARN in the form table/Books/index/*. - Granting Scan to callers that only need keyed reads. The AWS conditions page notes that Scan carries no LeadingKeys value, so a LeadingKeys condition needs a Null check to be effective. ## A read-only grant for one table lists three actions AWS publishes this statement for a Books table in us-west-2. For writes, AWS gives a separate policy that covers table Books and all its indexes with PutItem, UpdateItem, DeleteItem, BatchWriteItem, GetItem, BatchGetItem, Scan, Query and ConditionCheckItem. | Sid | Actions | Resource | | --- | --- | --- | | DescribeQueryScanBooksTable | dynamodb:DescribeTable, dynamodb:Query, dynamodb:Scan | arn:aws:dynamodb:us-west-2:111122223333:table/Books | | AccessTableAllIndexesOnBooks (write policy) | nine data actions listed above | arn:aws:dynamodb:us-west-2:123456789012:table/Books and table/Books/index/* | ## Condition keys limit access to items and attributes - dynamodb:LeadingKeys holds the partition key value of a request. AWS says to use the ForAllValues modifier and shows a policy tying it to the caller identity, for example `${www.amazon.com:user_id}`. - dynamodb:Attributes lists the attributes in a request, which restricts which columns can be read or written. AWS describes this as attribute-level control, alongside item-level control through key values and operation-specific rules. - dynamodb:Select represents the Select parameter of a request. - For BatchGetItem and BatchWriteItem, LeadingKeys contains the partition key of every item. For transactions, AWS says each item is authorized separately, using the action for that item such as dynamodb:PutItem or dynamodb:ConditionCheckItem. ## Check use with policy generation and CloudTrail IAM Access Analyzer can generate a policy from CloudTrail activity for up to 90 days. AWS says generated policies omit action-level activity for data events, so item reads and writes need CloudTrail data event logging or application logs to confirm. AWS also recommends MFA conditions on API calls where IAM users or the root user are required, and Access Analyzer validation before a policy ships. Compare the actions a role is granted against the table names and operations that appear in CloudTrail, and remove grants for tables the application no longer reads. ## Related pages - [Least privilege for AWS KMS](/services/aws-kms/) - [Least privilege for AWS Secrets Manager](/services/aws-secrets-manager/) - [Least privilege for Azure Storage](/services/azure-storage/) - [All of them](/services/) ## Sources - Identity-based policy examples for DynamoDB: https://docs.aws.amazon.com/amazondynamodb/latest/developerguide/security_iam_id-based-policy-examples.html (read 16 Nov 2026) - Using identity-based policies with DynamoDB: https://docs.aws.amazon.com/amazondynamodb/latest/developerguide/using-identity-based-policies.html (read 16 Nov 2026) - Policy for a table and its indexes: https://docs.aws.amazon.com/amazondynamodb/latest/developerguide/iam-policy-specific-table-indexes.html (read 16 Nov 2026) - IAM policy conditions for fine-grained access control: https://docs.aws.amazon.com/amazondynamodb/latest/developerguide/specifying-conditions.html (read 16 Nov 2026) - IAM Access Analyzer policy generation: https://docs.aws.amazon.com/IAM/latest/UserGuide/access-analyzer-policy-generation.html (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/services/aws-dynamodb/ --- # Least privilege for AWS KMS means separate key administrators and key users, with encryption context conditions Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. AWS KMS keys are governed by a key policy, and IAM policies work only if the key policy allows them. AWS documentation read on 16 Nov 2026 recommends separate administrator and user statements and encryption context conditions. This guide gives the key user action list and the checks. ## A key policy is required and IAM policies work only if it allows them AWS documentation says the default key policy contains an "Enable IAM User Permissions" statement for the account root principal. Without it, IAM policies that allow access to the key are ineffective, though IAM policies that deny access still work. The console also adds a key administrators statement and a key users statement. AWS says to give permissions in key policies where possible, rather than IAM policies that can apply to many keys, especially for kms:PutKeyPolicy and kms:ScheduleKeyDeletion. The identities that call this service (people, services, workloads, third parties) are covered in the [AWS service identities guide](/guides/aws-service-identities/) and the [workload identities guide](/guides/aws-workload-identities/). ## kms:* in both the key policy and an IAM policy gives full control - kms:* in the key policy and in an IAM policy. AWS says the principal then holds both administrative and usage permissions; delegate this only to named principals. - The default administrators list, which uses wildcards such as kms:Create*, kms:Put* and kms:Delete*. AWS says new API operations are then allowed automatically, and you can change the policy to a fixed set. - kms:TagResource and kms:UntagResource for administrators. AWS says tags can control access to the key, so these permissions can allow or deny access. - Unrestricted kms:CreateKey. AWS says creators set the key policy, so limit it, for example with the kms:KeySpec condition. ## A key user statement lists five actions AWS shows the statement the console adds for a symmetric encryption KMS key, naming a role as the principal. Users get the cryptographic operations but not the administrative ones. In the administrators statement, kms:CreateKey is not listed because AWS says it is valid only in an IAM policy, and kms:Delete* does not allow ScheduleKeyDeletion, which is its own entry. | Sid | Principal | Actions | Resource | | --- | --- | --- | --- | | Allow use of the key | arn:aws:iam::111122223333:role/ExampleKeyUserRole | kms:Decrypt, kms:DescribeKey, kms:Encrypt, kms:GenerateDataKey*, kms:ReEncrypt* | * | ## Encryption context and source conditions tie a grant to one resource AWS calls encryption context condition keys the most effective way to apply least privilege to cryptographic operations, because they bind a permission to the context attached to the ciphertext. Use kms:EncryptionContext:context-key or kms:EncryptionContextKeys only with CreateGrant or symmetric operations such as GenerateDataKey and Decrypt; AWS says they cause a deny if used for DescribeKey. For an AWS service principal, AWS strongly recommends aws:SourceArn or aws:SourceAccount. The Amazon Location example combines kms:ViaService, kms:CallerAccount and an encryption context for one tracker, so the permission to create grants applies only to that tracker. AWS says the encryption context for Secrets Manager identifies the secret and its version. ## Test in non-production and watch policy changes AWS recommends IAM Access Analyzer when testing an application outside production to apply least privilege to IAM policies. It also says to limit who can create and attach IAM policies and to monitor for policy changes if you manage key access through IAM. For IAM users, AWS strongly recommends MFA, including before privileged actions such as scheduling key deletion. ## Related pages - [Least privilege for AWS Secrets Manager](/services/aws-secrets-manager/) - [Least privilege for Azure Storage](/services/azure-storage/) - [Least privilege for Azure Key Vault](/services/azure-key-vault/) - [All of them](/services/) ## Sources - Least-privilege permissions for KMS: https://docs.aws.amazon.com/kms/latest/developerguide/least-privilege.html (read 16 Nov 2026) - Default key policy: https://docs.aws.amazon.com/kms/latest/developerguide/key-policy-default.html (read 16 Nov 2026) - Best practices for IAM policies in KMS: https://docs.aws.amazon.com/kms/latest/developerguide/iam-policies-best-practices.html (read 16 Nov 2026) - KMS condition keys: https://docs.aws.amazon.com/kms/latest/developerguide/conditions-kms.html (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/services/aws-kms/ --- # Least privilege for AWS Secrets Manager names each secret ARN and its KMS key Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. AWS Secrets Manager uses IAM identity-based and resource-based policies, plus the KMS key that encrypts each secret. AWS documentation read on 16 Nov 2026 warns against giving end users the administrator pair. This guide shows a one-secret read grant, wildcard options and what to review. ## Secrets Manager access combines IAM policies and KMS permissions AWS documentation says Secrets Manager uses IAM policies to define who can reach which secrets and what they can do. You can attach identity-based policies to users, groups and roles, or resource-based policies to secrets, and you can use attribute-based access control. Secrets are encrypted with AWS KMS. The AWS managed key aws/secretsmanager has the correct permissions automatically; a different KMS key needs its own permissions. Rotation functions are Lambda functions that need access to the secret and the target database or service. AWS lists separate permission topics for administrators, access to secrets, rotation functions, encryption keys and replication, and says an application on EC2 can reach a secret through a role on its instance profile. Other services that take roles include Amazon Redshift, AWS Lambda and Amazon ECS. The identities that call this service (people, services, workloads, third parties) are covered in the [AWS service identities guide](/guides/aws-service-identities/) and the [workload identities guide](/guides/aws-workload-identities/). ## The administrator pair is not meant for end users - SecretsManagerReadWrite together with IAMFullAccess is the documented administrator combination. - AWS says not to grant administrator permissions to end users, because the permission needed to enable rotation, IAMFullAccess, grants significant permissions that are not appropriate for them. - A Resource of "*" with secretsmanager:GetSecretValue. AWS example policies name the secret ARN instead. - Batch grants without limits. AWS shows BatchGetSecretValue with ListSecrets on "*" paired with a statement naming the allowed secret ARNs, so a batch call cannot return others. ## Reading one secret needs the secret ARN and the key For a secret encrypted with a customer managed key, AWS shows two statements. Add secretsmanager:DescribeSecret if the caller also needs metadata. | Action | Resource | Why | | --- | --- | --- | | secretsmanager:GetSecretValue | arn:aws:secretsmanager:us-east-1:123456789012:secret:secretName-AbCdEf | Read the value of one secret | | kms:Decrypt | arn:aws:kms:us-east-1:123456789012:key/key-id | Decrypt with the customer managed key | | secretsmanager:DescribeSecret (optional) | the same secret ARN | Read metadata without the value | ## Wildcards, ABAC, key denies and replication controls narrow scope - Wildcards: AWS notes that secret ARNs end in six random characters, so a name followed by the six-character wildcard "??????" matches one secret name exactly. - A deny on a specific KMS key stops secrets being encrypted with it. - Attribute-based access control lets tags decide access. - AWS documents a policy to prevent replication to other Regions and a separate set of permissions for cross-account replication. - The KMS encryption context for Secrets Manager identifies the secret and its version, so a key policy can bind to it. ## Review who can reach each secret The AWS access control page includes a topic on determining who has permissions to your secrets, and another on reaching secrets from a different account or an on-premises environment. The page ends with topics on identity-based policies and resource-based policies, so both can be reviewed. Compare the principals that appear with the roles that call GetSecretValue in your logs, and remove the rest. For a secret read by one application, the grant above is the whole requirement; add DescribeSecret only where code reads metadata. ## Related pages - [Least privilege for Azure Storage](/services/azure-storage/) - [Least privilege for Azure Key Vault](/services/azure-key-vault/) - [Least privilege for Azure Kubernetes Service](/services/azure-aks/) - [All of them](/services/) ## Sources - Authentication and access control for Secrets Manager: https://docs.aws.amazon.com/secretsmanager/latest/userguide/auth-and-access.html (read 16 Nov 2026) - Identity-based policies for Secrets Manager: https://docs.aws.amazon.com/secretsmanager/latest/userguide/auth-and-access_examples.html (read 16 Nov 2026) - Least-privilege permissions for KMS: https://docs.aws.amazon.com/kms/latest/developerguide/least-privilege.html (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/services/aws-secrets-manager/ --- # Least privilege for Azure Storage means a data role at container scope, with Shared Key off Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. Azure Storage separates management roles from data roles. Microsoft's documentation, read 16 Nov 2026, says data access needs a data role such as Storage Blob Data Reader, which can be scoped to one container and narrowed further with role assignment conditions. ## Azure Storage authorizes blob data through data roles, not management roles Microsoft's documentation says only roles explicitly defined for data access permit a security principal to read blob data through Microsoft Entra ID. Built-in roles such as Owner, Contributor and Storage Account Contributor manage a storage account but do not grant blob data access by themselves. There is one exception that matters for least privilege. Microsoft's documentation says a role that includes Microsoft.Storage/storageAccounts/listKeys/action can still reach the data, because the account key opens it through Shared Key authorization. Role assignments can be placed at five levels: a single container, the storage account, the resource group, the subscription and a management group. Microsoft says always to grant the narrowest possible scope, and that roles at broader scopes are inherited by everything beneath them. The identities that call this service (people, services, workloads, third parties) are covered in the [Azure service identities guide](/guides/azure-service-identities/) and the [workload identities guide](/guides/azure-workload-identities/). ## Roles and settings to avoid, with the documented reason Four choices widen access more than most tasks need. - Storage Account Contributor: Microsoft's role description says it provides access to the account key, which can be used to access data via Shared Key authorization. - Storage Blob Data Owner: Microsoft describes it as full access to blob containers and data, including assigning POSIX access control. Use it only where ownership or POSIX control is the job. - Storage Blob Data Contributor at account, resource group or subscription scope: it reads, writes and deletes, and the scope decides how many containers that covers. - Shared Key authorization: Microsoft's documentation says it is not recommended because it may be less secure, and advises disabling it for the storage account. ## A minimal binding for read-only access to one container For a reporting job that reads files from a single container, assign the Storage Blob Data Reader role at the container scope. The role ID and permissions below come from Microsoft's built-in role listing. | Item | Value | | --- | --- | | Role | Storage Blob Data Reader | | Role ID | 2a2b9908-6ea1-4ae2-8e65-a410df84e7d1 | | Management actions | Microsoft.Storage/storageAccounts/blobServices/containers/read; Microsoft.Storage/storageAccounts/blobServices/generateUserDelegationKey/action | | Data action | Microsoft.Storage/storageAccounts/blobServices/containers/blobs/read | | Scope | The one container, not the account | ## Storage's own controls narrow access further Role assignment conditions add an attribute check to a data role. Microsoft's documentation says conditions can be added to built-in roles including Storage Blob Data Reader, Contributor and Owner, and that they are evaluated when the action is authorized. Microsoft gives one example: read access to blobs only if the blobs are tagged Project=Sierra. It also states the trade-off: conditions need a consistent attribute taxonomy, and the attributes must be protected so that access is not compromised. Conditions are not supported for management actions through the Storage resource provider. For shared access signatures, Microsoft recommends a user delegation SAS, secured with Microsoft Entra credentials rather than the account key. The Storage Blob Delegator role is the one that grants the delegation key. ## Checking what is used Microsoft Defender for Cloud permissions management, per Microsoft's documentation, continuously analyzes identity configurations and usage patterns to identify excessive, unused or misconfigured permissions, including for managed identities and service principals. Microsoft Entra access reviews can be scheduled weekly, monthly, quarterly or annually so that owners confirm or remove role assignments. Microsoft's role-based access control best practices also say to limit subscription owners to a maximum of 3 and to avoid privileged administrator roles when a job function role fits. For storage, list role assignments on the account and ask of each one whether the holder reads, writes or deletes, and at which container. ## Related pages - [Least privilege for Azure Key Vault](/services/azure-key-vault/) - [Least privilege for Azure Kubernetes Service](/services/azure-aks/) - [Least privilege for Azure Functions](/services/azure-functions/) - [All of them](/services/) ## Sources - Azure built-in roles for storage: https://learn.microsoft.com/en-us/azure/role-based-access-control/built-in-roles/storage (read 16 Nov 2026) - Authorize access to blob data with Microsoft Entra ID: https://learn.microsoft.com/en-us/azure/storage/blobs/authorize-access-azure-active-directory (read 16 Nov 2026) - Azure role assignment conditions for Blob Storage: https://learn.microsoft.com/en-us/azure/storage/blobs/storage-auth-abac (read 16 Nov 2026) - Azure RBAC best practices: https://learn.microsoft.com/en-us/azure/role-based-access-control/best-practices (read 16 Nov 2026) - Defender for Cloud permissions management: https://learn.microsoft.com/en-us/azure/defender-for-cloud/permissions-management (read 16 Nov 2026) - Microsoft Entra access reviews: https://learn.microsoft.com/en-us/entra/id-governance/access-reviews-overview (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/services/azure-storage/ --- # Least privilege for Azure Key Vault means one vault per application and environment, with Secrets User or Crypto User Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. Key Vault has a management plane and a data plane that are authorized separately. Microsoft's documentation, read 16 Nov 2026, recommends one vault per application per environment with data plane roles assigned at the vault scope, and names Key Vault Secrets User for reading secrets. ## Key Vault authorizes its management plane and data plane separately Microsoft's documentation says both planes use Microsoft Entra ID for authentication. The control plane uses Azure RBAC for authorization. The data plane uses either a Key Vault access policy (the legacy model) or Azure RBAC for data plane operations. The two planes work independently. A user can hold read access to vault properties and tags without any access to keys, secrets or certificates. Changing a vault from access policies to the Azure RBAC permission model needs unrestricted Microsoft.Authorization/roleAssignments/write, which is part of the Owner and User Access Administrator roles. Microsoft recommends a vault per application per environment (development, pre-production, production), with roles assigned at the key vault scope. The identities that call this service (people, services, workloads, third parties) are covered in the [Azure service identities guide](/guides/azure-service-identities/) and the [workload identities guide](/guides/azure-workload-identities/). ## Roles and patterns to avoid, with the documented reason Most over-grants in Key Vault come from the control plane reaching into the data plane. - Contributor on a vault that still uses access policies: Microsoft's documentation says a user with Contributor on the control plane can grant themselves data plane access by setting an access policy, and that Contributor access to key vaults should be tightly controlled. - Key Vault Administrator: Microsoft describes it as all data plane operations on a vault and all objects in it. Reserve it for few people. - Roles on individual keys, secrets or certificates: Microsoft says this is not recommended, except where individual secrets need individual user access or must be shared between applications. - Object-scope assignments as tenant isolation: Microsoft's FAQ says administrative operations such as network access control and monitoring need vault-level permissions, so object-level assignment does not isolate application teams. ## A minimal binding for an application that reads secrets An application that reads secret values at start-up needs Key Vault Secrets User on its own vault. Role names and IDs come from Microsoft's Key Vault RBAC guide. | Task | Built-in role | Role ID | What Microsoft says it allows | | --- | --- | --- | --- | | Read secret values | Key Vault Secrets User | 4633458b-17de-408a-b874-0445c86b69e6 | Read secret contents, including the secret portion of a certificate with private key | | Use a key for cryptographic operations | Key Vault Crypto User | 12338af0-0e69-4776-bea7-57ae8d297424 | Perform cryptographic operations using keys | | Wrap and unwrap for storage encryption | Key Vault Crypto Service Encryption User | e147488a-f6f5-4113-8e2d-b22465e65bf6 | Read metadata of keys and perform wrap and unwrap operations | | Read metadata only | Key Vault Reader | 21090545-7ca7-4776-b22c-e363652d74d2 | Cannot read secret values or key material | ## Vault scope and the permission model are the narrowing controls Assign the role at the vault scope, using a command of the form az role assignment create --role "Key Vault Secrets User" --assignee --scope . Use a managed identity for the application so that no secret is needed to reach the vault. Microsoft's guide notes that a custom role can be defined when no predefined role fits, and that the Key Vault Contributor role is for control plane operations only: it does not allow access to keys, secrets or certificates. Officer roles (Secrets Officer, Crypto Officer, Certificates Officer) manage objects but not permissions. ## Checking what is used Defender for Cloud permissions management, according to Microsoft, assesses managed identities, service principals and groups for excessive or unused permissions. Run it against the identities that hold vault roles, then remove Secrets Officer or Administrator assignments held by identities that only read. For people with administrative roles, Microsoft Entra access reviews let owners confirm each assignment on a schedule. List role assignments at the vault scope with az role assignment list and compare them with the table above. ## Related pages - [Least privilege for Azure Kubernetes Service](/services/azure-aks/) - [Least privilege for Azure Functions](/services/azure-functions/) - [Least privilege for Azure SQL Database](/services/azure-sql-database/) - [All of them](/services/) ## Sources - Azure RBAC guide for Key Vault: https://learn.microsoft.com/en-us/azure/key-vault/general/rbac-guide (read 16 Nov 2026) - Defender for Cloud permissions management: https://learn.microsoft.com/en-us/azure/defender-for-cloud/permissions-management (read 16 Nov 2026) - Microsoft Entra access reviews: https://learn.microsoft.com/en-us/entra/id-governance/access-reviews-overview (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/services/azure-key-vault/ --- # Least privilege for AKS means Azure RBAC roles at namespace scope, not Cluster Admin Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. AKS can authorize Kubernetes API calls with Azure RBAC. Microsoft's documentation, read 16 Nov 2026, lists four built-in roles, advises cluster or namespace scope, and warns that the Writer role can reach Secrets and run pods as any service account in the namespace. ## AKS can authorize Kubernetes API calls with Azure RBAC role assignments Microsoft's documentation says Microsoft Entra ID authorization for the Kubernetes API uses Azure RBAC role assignments to grant access to Kubernetes resources. For built-in Kubernetes resources, assign one of the AKS built-in roles at the cluster or namespace scope. For custom resources, assign a custom role with Azure ABAC conditions that name the CRD groups or kinds. Microsoft also notes a split: with integrated Entra authentication, users and groups can be subjects in Kubernetes RBAC, but Entra role assignments and Kubernetes RBAC bindings are then managed separately. Choosing one model for Kubernetes API access avoids having two places to audit. Azure RBAC for the Azure resource itself (the cluster in the portal and the Azure Resource Manager API) is a different layer from Kubernetes API authorization. A person can hold one without the other. The identities that call this service (people, services, workloads, third parties) are covered in the [Azure service identities guide](/guides/azure-service-identities/) and the [workload identities guide](/guides/azure-workload-identities/). ## Roles to avoid, with the documented reason Three of the four built-in roles carry a documented caution. - Azure Kubernetes Service RBAC Cluster Admin: Microsoft's description is super-user access to perform any action on any resource, with full control over every resource in the cluster and all namespaces. - Azure Kubernetes Service RBAC Writer: it cannot view or modify roles or role bindings, but Microsoft says it allows accessing Secrets and running pods as any service account in the namespace, so it can gain the API access of any service account there. - Azure Kubernetes Service RBAC Admin: intended to be granted within a namespace, it can create roles and role bindings in that namespace. Microsoft's description says it does not allow write access to resource quota or to the namespace itself. - Azure Kubernetes Service RBAC Reader is the lowest of the four, and Microsoft says it cannot view Secrets because reading them would allow API access as any service account in the namespace. ## A minimal binding for a team that reads one namespace Microsoft's documentation shows the Reader and Writer role assignments scoped to a namespace by appending /namespaces/ to the cluster resource ID. | Need | Built-in role | Scope | | --- | --- | --- | | Read most objects in one namespace | Azure Kubernetes Service RBAC Reader | $AKS_ID/namespaces/ | | Deploy workloads in one namespace | Azure Kubernetes Service RBAC Writer | $AKS_ID/namespaces/ | | Manage roles and bindings in one namespace | Azure Kubernetes Service RBAC Admin | $AKS_ID/namespaces/ | | Any action anywhere in the cluster | Azure Kubernetes Service RBAC Cluster Admin | $AKS_ID (few people) | ## Namespace scope, custom roles and ABAC conditions narrow the rest Microsoft's article shows custom role definitions, for example an AKS Deployment Reader role created from a JSON file and assigned with az role assignment create. Use a custom role when the four built-in roles cover more verbs than the task needs. A practical note from the documentation: after you create a namespace-scoped assignment, the portal does not show it at that scope. Find it with az role assignment list --scope $AKS_ID/namespaces/, or list assignments for the user or group. ## Checking what is used Run the list command above for every namespace and for the cluster, then compare it with who deploys, who debugs and who only reads. Microsoft Entra access reviews can recur weekly to annually for the groups that hold these roles. Defender for Cloud permissions management, per Microsoft's documentation, flags excessive and unused permissions for identities including managed identities and service principals, which covers the identities that automation uses to reach the cluster. ## Related pages - [Least privilege for Azure Functions](/services/azure-functions/) - [Least privilege for Azure SQL Database](/services/azure-sql-database/) - [Least privilege for Cloud Storage](/services/gcp-cloud-storage/) - [All of them](/services/) ## Sources - Azure RBAC for Kubernetes authorization in AKS: https://learn.microsoft.com/en-us/azure/aks/manage-azure-rbac (read 16 Nov 2026) - Microsoft Entra access reviews: https://learn.microsoft.com/en-us/entra/id-governance/access-reviews-overview (read 16 Nov 2026) - Defender for Cloud permissions management: https://learn.microsoft.com/en-us/azure/defender-for-cloud/permissions-management (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/services/azure-aks/ --- # Least privilege for Azure Functions means managed identities, Website Contributor and no master key Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. Azure Functions supports only Contributor, Owner and Reader at the function app level. Microsoft's documentation, read 16 Nov 2026, advises running the app with the lowest possible permissions, using managed identities for connections, and disabling the administrative endpoints that require the app's master key. ## Azure Functions has two permission surfaces: who manages the app and what the app can reach Microsoft's documentation says Functions supports Azure RBAC with three supported roles: Contributor, Owner and Reader. Permissions take effect at the function app level. Contributor is required for most function app tasks, and only Owner can delete a function app. The second surface is the identity the app runs as. Microsoft says a managed identity from Microsoft Entra ID lets the app reach other protected resources, such as Key Vault, with the platform managing the identity so there is no secret to provision or rotate. Microsoft's advice for this second surface is plain: as with any application or service, run your function app with the lowest possible permissions. The identities that call this service (people, services, workloads, third parties) are covered in the [Azure service identities guide](/guides/azure-service-identities/) and the [workload identities guide](/guides/azure-workload-identities/). ## Settings and roles to avoid, with the documented reason The documented risks sit in keys, shared credentials and wide roles. - The master key on exposed /admin endpoints: Microsoft says administrative endpoints require the app's master key when exposed, and describes setting the functionsRuntimeAdminIsolationEnabled site property to disable them. - Contributor on a function app for people who only read logs: Microsoft says log data in Application Insights needs Contributor together with the Monitoring Reader permission, so check whether Reader plus Monitoring Reader meets the need. - Connection strings in app settings: Microsoft says identity-based connections give more fine-grained access control and auditing than secrets, and describes Key Vault references as the alternative where a secret must exist. - Wildcards in the CORS allowed origins list: Microsoft says to list the specific domains instead. ## A minimal binding for a person who publishes one app Microsoft's documentation says a user needs at least Contributor level permissions on an app, including the built-in Website Contributor role, to be granted access to app-level deployment credentials through RBAC. Reader cannot publish. | Item | Value | | --- | --- | | Role | Website Contributor | | Microsoft's description | Manage websites, but not web plans. Does not allow you to assign roles in Azure RBAC. | | Scope | The single function app or its resource group | | App identity | A managed identity granted a data role on each dependency, for example Key Vault Secrets User on one vault | ## Function-level controls narrow what an attacker or a mistake can reach Microsoft lists several controls. Use App Service Authentication, or Azure API Management in front of the app, so that callers are authenticated and authorized by identity. Restrict access to the APIM instance's IP address. Run the app inside a virtual network with private endpoints for the highest level of security. Set a daily GB-sec usage quota on Consumption plan apps to limit execution if malicious code runs. Application-scope deployment credentials can deploy only that app and cannot be configured manually, while user-scope credentials work across every app the user can access. ## Checking what is used Microsoft says Functions integrates with Application Insights and with Azure Monitor Logs through diagnostic settings, which can stream platform logs to a Log Analytics workspace. Use those logs to list which connections the app really calls, then remove roles for the rest. Defender for Cloud permissions management reports excessive and unused permissions for managed identities, according to Microsoft's documentation. Review the identity of each function app with it before you widen any role. ## Related pages - [Least privilege for Azure SQL Database](/services/azure-sql-database/) - [Least privilege for Cloud Storage](/services/gcp-cloud-storage/) - [Least privilege for BigQuery](/services/gcp-bigquery/) - [All of them](/services/) ## Sources - Securing Azure Functions: https://learn.microsoft.com/en-us/azure/azure-functions/security-concepts (read 16 Nov 2026) - Azure built-in roles for web and mobile: https://learn.microsoft.com/en-us/azure/role-based-access-control/built-in-roles/web-and-mobile (read 16 Nov 2026) - Defender for Cloud permissions management: https://learn.microsoft.com/en-us/azure/defender-for-cloud/permissions-management (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/services/azure-functions/ --- # Least privilege for Azure SQL Database means custom database roles for Entra groups, not db_owner Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. Azure SQL Database has two layers: Azure RBAC for managing servers and databases, and database roles for data. Microsoft's documentation, read 16 Nov 2026, advises granting the least privileges necessary and says the db_datareader fixed role grants read access to every table in the database. ## Azure SQL Database separates resource management from data permissions Microsoft's documentation says Azure RBAC roles such as SQL DB Contributor and SQL Security Manager control management of SQL resources, while authorization to data is managed through database roles and explicit permissions. A management role does not by itself carry SQL query rights. Microsoft describes the server admin created at deployment as mapping to the dbo user in each database, with all database permissions as a member of db_owner. It also describes a Microsoft Entra administrator, which can be an individual or a security group, with full administrative permissions. Microsoft's guidance is to grant users the least privileges necessary, and to assign permissions to Entra security groups and roles instead of to individual users. The identities that call this service (people, services, workloads, third parties) are covered in the [Azure service identities guide](/guides/azure-service-identities/) and the [workload identities guide](/guides/azure-workload-identities/). ## Roles to avoid, with the documented reason Several roles look administrative-only but reach data indirectly. - SQL Server Contributor: Microsoft says it grants no SQL data-plane permissions directly, but control of administrator credentials can enable data-plane access. - SQL DB Contributor: Microsoft says it grants no SQL query permissions directly, but its data-movement capabilities, such as copy, restore and export, can enable access. - db_owner: Microsoft says to use it to grant full permission to only a few users. - db_datareader: Microsoft says it grants read access to every table in the database, which is more than is strictly necessary, and that the fixed roles are not recommended for most production databases. - ##MS_DatabaseManager## and ##MS_LoginManager## membership: Microsoft says members can create and manage databases or logins, and a member who creates a database maps to db_owner in it. ## A minimal pattern for a reporting group Microsoft's pattern is to put Entra users into a security group, create a contained database user for the group, and add that user to a custom role with specific permissions. | Step | Statement or setting | Source detail | | --- | --- | --- | | 1. Group | An Entra security group for the reporting team | Microsoft: use groups, not individual users | | 2. Database user | A contained user mapped to the group | Created while signed in with an Entra account | | 3. Role | CREATE ROLE, then grant only the needed object permissions | Microsoft: carefully grant each role the least permissions necessary | | 4. Membership | ALTER ROLE ... ADD MEMBER for the group user | Permissions aggregate across all roles a user belongs to | ## Narrowing controls inside the service Fixed roles with a deny form exist: Microsoft lists db_denydatawriter and db_denydatareader among the nine fixed database roles. Custom roles remain the main tool, because they can be granted on specific objects instead of the whole database. Microsoft notes that when a user belongs to several roles, the permissions add up, so removing a role does not help if another role grants the same right. Use the Microsoft Entra administrator as a group, so that membership changes do not require touching the server. ## Checking what is used Defender for Cloud permissions management covers identities that hold Azure RBAC roles on SQL resources, flagging excessive and unused permissions according to Microsoft's documentation. Entra access reviews can recertify membership of the groups that map to database users. Inside the database, list role memberships and object permissions and compare them with the tasks each group performs. Remove any member of db_owner who has not changed schema. ## Related pages - [Least privilege for Cloud Storage](/services/gcp-cloud-storage/) - [Least privilege for BigQuery](/services/gcp-bigquery/) - [Least privilege for Google Kubernetes Engine](/services/gcp-gke/) - [All of them](/services/) ## Sources - Azure built-in roles for databases: https://learn.microsoft.com/en-us/azure/role-based-access-control/built-in-roles/databases (read 16 Nov 2026) - Azure SQL logins and users: https://learn.microsoft.com/en-us/azure/azure-sql/database/logins-create-manage (read 16 Nov 2026) - Defender for Cloud permissions management: https://learn.microsoft.com/en-us/azure/defender-for-cloud/permissions-management (read 16 Nov 2026) - Microsoft Entra access reviews: https://learn.microsoft.com/en-us/entra/id-governance/access-reviews-overview (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/services/azure-sql-database/ --- # Least privilege for Cloud Storage means Storage Object Viewer on one bucket, with uniform bucket-level access Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. Cloud Storage grants access through IAM roles on projects, buckets or managed folders. Google's documentation, read 16 Nov 2026, lists Storage Object Viewer, Creator, User and Admin, warns that basic roles confer legacy bucket roles, and ties IAM Conditions to uniform bucket-level access. ## Cloud Storage permissions come from IAM roles at project, bucket or managed folder level Google's documentation says the Cloud Storage roles can be applied to projects, buckets or managed folders, except the legacy roles, which can be granted only on individual buckets. A role applied to a bucket controls only that bucket and the objects in it. Four predefined roles cover most work. Storage Object Viewer (roles/storage.objectViewer) views objects and metadata, excluding ACLs, and lists objects. Storage Object Creator (roles/storage.objectCreator) creates objects but cannot view, delete or overwrite them. Storage Object User (roles/storage.objectUser) creates, reads, updates and deletes objects. Storage Admin (roles/storage.admin) has full control of objects and buckets. The identities that call this service (people, services, workloads, third parties) are covered in the [Google Cloud service identities guide](/guides/google-cloud-service-identities/) and the [workload identities guide](/guides/google-cloud-workload-identities/). ## Roles to avoid, with the documented reason Basic roles and legacy roles bring more than the name suggests. - Basic roles Owner, Editor and Viewer: Google's documentation says principals granted them gain legacy bucket roles through convenience values. Viewer gains Storage Legacy Bucket Reader and Legacy Object Reader. Editor and Owner gain Legacy Bucket Owner and Legacy Object Owner on the project's buckets. - Storage Admin at project scope: Google describes full control of objects and buckets, and applied to a project it covers every bucket. - Storage Object User where read is enough: it adds create, update, delete and multipart upload permissions that a reader never exercises. - Google's IAM overview adds that basic roles include thousands of permissions across all services and should not be granted in production unless there is no alternative. ## A minimal binding for a job that uploads files to one bucket A producer that writes new files and never reads or overwrites needs only the creator role, on the bucket. | Task | Predefined role | Level | Documented effect | | --- | --- | --- | --- | | Upload new objects only | roles/storage.objectCreator | Bucket | Create objects; no view, delete or overwrite | | Read and list objects | roles/storage.objectViewer | Bucket or managed folder | storage.objects.get and storage.objects.list | | Read, write and delete objects | roles/storage.objectUser | Bucket | Create, read, update, delete objects and multipart uploads | | Manage buckets and objects | roles/storage.admin | Bucket (few people) | Full control | ## Uniform bucket-level access, managed folders and IAM Conditions narrow access Google's IAM Conditions page says that to use conditions in the allow policy of a bucket you must enable uniform bucket-level access; where you cannot, add the conditional grant at a higher level such as the project. Its example expression scopes access with resource.name.startsWith('projects/_/buckets/exampleco-site-assets/'). Managed folders let a role be granted below the bucket for a prefix of objects. Remove the convenience-value access that basic roles confer by editing the bucket's IAM policy once the bucket exists, as Google's documentation describes. ## Checking what is used Google's IAM recommender generates policy insights by comparing a principal's permissions with those used in the last 90 days, and suggests removing or replacing roles that give excess permissions. If the role was granted fewer than 90 days ago, it looks at usage since the grant. Review recommendations for storage roles at project scope, then repeat at bucket level for the buckets that hold sensitive data. ## Related pages - [Least privilege for BigQuery](/services/gcp-bigquery/) - [Least privilege for Google Kubernetes Engine](/services/gcp-gke/) - [Least privilege for Cloud Run](/services/gcp-cloud-run/) - [All of them](/services/) ## Sources - IAM roles for Cloud Storage: https://cloud.google.com/storage/docs/access-control/iam-roles (read 16 Nov 2026) - IAM roles overview: https://cloud.google.com/iam/docs/roles-overview (read 16 Nov 2026) - IAM Conditions overview: https://cloud.google.com/iam/docs/conditions-overview (read 16 Nov 2026) - Role recommendations overview: https://cloud.google.com/iam/docs/recommender-overview (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/services/gcp-cloud-storage/ --- # Least privilege for BigQuery means Job User on the project and Data Viewer on the dataset Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. BigQuery separates the right to run jobs from the right to read data. Google's documentation, read 16 Nov 2026, describes BigQuery Job User at project level and Data Viewer at dataset, table or view level, and says BigQuery Admin is typically granted at the project and is not recommended on datasets. ## BigQuery splits running jobs from reading data, and roles attach at several levels Google's documentation says BigQuery Job User (roles/bigquery.jobUser) provides permissions to run jobs, including queries, and can be granted only on Resource Manager resources: projects, folders and organizations. BigQuery Data Viewer (roles/bigquery.dataViewer) reads table data, and can be granted on a dataset, table, view or routine. That split is the basis of least privilege here. A person who must query one dataset needs Job User on the billing project and Data Viewer on that dataset, not a project-wide data role. Dataset-level grants map to the older dataset roles. Google notes Data Viewer maps to the READER basic role, so granting it on a dataset gives READER access to the dataset. The identities that call this service (people, services, workloads, third parties) are covered in the [Google Cloud service identities guide](/guides/google-cloud-service-identities/) and the [workload identities guide](/guides/google-cloud-workload-identities/). ## Roles to avoid, with the documented reason Broad BigQuery roles combine data, job and dataset management. - BigQuery Admin (roles/bigquery.admin): Google's description is permissions to manage all resources in the project, all data in it, and cancel other users' jobs. Google says it is possible to grant at lower levels but not recommended, because other predefined roles grant full permissions over those resources and are less permissive. - BigQuery User (roles/bigquery.user) at project level: it lets the holder run jobs, enumerate datasets, and create new datasets, becoming BigQuery Data Owner on any dataset it creates. - BigQuery Data Owner (roles/bigquery.dataOwner) and Data Editor (roles/bigquery.dataEditor): both write data; give them to pipelines that load tables, not to analysts. - Basic roles: Google's IAM overview says they are highly permissive and should be avoided in production unless there is no alternative. ## A minimal binding for an analyst who queries one dataset Two bindings, in two places, give a person query access to a single dataset. | Binding | Role | Resource | Documented effect | | --- | --- | --- | --- | | 1 | roles/bigquery.jobUser | The project that runs queries | Run jobs, including queries, in the project | | 2 | roles/bigquery.dataViewer | The dataset | List tables; get (query), export and replicate table data; invoke routines | ## Dataset, table, column and row controls narrow access further Google's access control documentation lists controls beyond roles: IAM conditions on access, authorized views, column-level access control with data policies, and row-level security with row-level access policies. Use authorized views to expose a subset of a table without granting access to the underlying table. Roles granted at a table or view carry less than the same role granted at the dataset. ## Checking what is used The IAM recommender compares each principal's roles with permissions used in the last 90 days and flags excess ones, according to Google's documentation. Run it against the project that hosts datasets and review roles such as BigQuery Admin and Data Editor first. Then list the access entries on each dataset and ask which of the principals have queried it recently. Google's IAM roles overview describes custom roles and predefined roles as the tools for matching a role to a task. For BigQuery, compare the permissions of a predefined role with the job: an analyst who only queries needs table reads and job creation, and does not need dataset creation or table deletion. Where no predefined role fits, a custom role can carry exactly the permissions the task uses. ## Related pages - [Least privilege for Google Kubernetes Engine](/services/gcp-gke/) - [Least privilege for Cloud Run](/services/gcp-cloud-run/) - [Least privilege for Cloud Functions](/services/gcp-cloud-functions/) - [All of them](/services/) ## Sources - BigQuery IAM roles and permissions: https://cloud.google.com/bigquery/docs/access-control (read 16 Nov 2026) - IAM roles overview: https://cloud.google.com/iam/docs/roles-overview (read 16 Nov 2026) - IAM Conditions overview: https://cloud.google.com/iam/docs/conditions-overview (read 16 Nov 2026) - Role recommendations overview: https://cloud.google.com/iam/docs/recommender-overview (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/services/gcp-bigquery/ --- # Least privilege for GKE means Kubernetes RBAC inside clusters and IAM roles only for cluster management Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. GKE access is checked by IAM at the project level and by Kubernetes RBAC at cluster and namespace level. Google's documentation, read 16 Nov 2026, says an entity needs enough of both, and names the Kubernetes Engine Default Node Service Account role as the least privilege role for nodes. ## GKE combines IAM at project level with Kubernetes RBAC at cluster and namespace level Google's documentation says RBAC controls access on a cluster and namespace level, while IAM works on the project level, and that an entity must have sufficient RBAC and IAM permissions to work with resources in a cluster. Predefined GKE roles include Kubernetes Engine Admin (roles/container.admin), Kubernetes Engine Developer (roles/container.developer), Kubernetes Engine Viewer (roles/container.viewer) and Kubernetes Engine Cluster Viewer (roles/container.clusterViewer). The last gets and lists clusters and includes container.clusters.connect. Nodes run as a service account. Google documents roles/container.defaultNodeServiceAccount as the least privilege role to use as the default service account for GKE nodes. The identities that call this service (people, services, workloads, third parties) are covered in the [Google Cloud service identities guide](/guides/google-cloud-service-identities/) and the [workload identities guide](/guides/google-cloud-workload-identities/). ## Roles to avoid, with the documented reason Four patterns widen access more than a task needs. - Kubernetes Engine Admin: Google's description is full management of clusters and their Kubernetes API objects. It is a project-level role, so it covers every cluster in the project. - Kubernetes Engine Developer at project level: Google describes access to Kubernetes API objects inside clusters, which is a long list of container.* permissions for all clusters in the project. - Setting a node service account: Google says you must also have the Service Account User role (roles/iam.serviceAccountUser) on the user-managed service account the nodes will use. Treat that grant as a way to act as the account. - Basic roles on the project: Google says they are highly permissive and should be avoided in production unless there is no alternative. ## A minimal binding for a developer who connects to clusters Google's role listing offers a connect-only role. Pair it with a namespace-scoped Kubernetes RBAC binding so that IAM stays narrow and the real permissions are set where Kubernetes enforces them. | Layer | Role | What Google says it provides | | --- | --- | --- | | IAM, project | roles/container.clusterViewer | Get and list GKE clusters; includes container.clusters.connect | | IAM, project | roles/container.viewer | Read-only access to resources in clusters, such as nodes, pods and GKE API objects | | IAM, project | roles/container.developer | Access to Kubernetes API objects inside clusters | | Kubernetes RBAC | A Role and RoleBinding in one namespace | Set per namespace by the cluster owner | ## RBAC, Google Groups and node account choice narrow access Google's documentation has a section on authorizing actions in clusters with GKE RBAC and one on managing permissions for groups using Google Groups with RBAC. Binding groups, rather than individuals, makes a leaver's access disappear with one membership change. Use the Default Node Service Account role for nodes and grant extra access to workloads through separate identities, not through the node account, since every pod on the node shares that account. ## Checking what is used The IAM recommender flags excess permissions on IAM roles using the last 90 days of usage, per Google's documentation. It covers the project-level GKE roles above and the roles of the node service accounts. For Kubernetes RBAC, list RoleBindings and ClusterRoleBindings per namespace and compare them with the groups that actually deploy. Remove cluster-wide bindings that exist for one namespace. ## Related pages - [Least privilege for Cloud Run](/services/gcp-cloud-run/) - [Least privilege for Cloud Functions](/services/gcp-cloud-functions/) - [Least privilege for Cloud KMS](/services/gcp-cloud-kms/) - [All of them](/services/) ## Sources - GKE IAM roles: https://cloud.google.com/kubernetes-engine/docs/how-to/iam (read 16 Nov 2026) - IAM roles overview: https://cloud.google.com/iam/docs/roles-overview (read 16 Nov 2026) - Role recommendations overview: https://cloud.google.com/iam/docs/recommender-overview (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/services/gcp-gke/ --- # Least privilege for Cloud Run means a dedicated service identity, Cloud Run Invoker per service and Developer for deploys Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. Cloud Run splits deploying a service from invoking it. Google's documentation, read 16 Nov 2026, says deploying needs Cloud Run Developer, Artifact Registry Reader and Service Account User on the service identity, and warns that Service Account User lets a holder act as that identity. ## Cloud Run has three separate grants: deploy, invoke and run-as Google's documentation lists four predefined roles. Cloud Run Admin (roles/run.admin) gives full control over all Cloud Run resources. Cloud Run Developer (roles/run.developer) gives read and write access to all Cloud Run resources. Cloud Run Viewer (roles/run.viewer) views the state of all resources, including IAM policies. Cloud Run Invoker (roles/run.invoker) invokes services and instances and runs jobs. All four can be granted on a Cloud Run service or job, not only on the project, so a deployer or invoker can be limited to one service. The third grant is the identity the service runs as. Deploying with a service identity needs iam.serviceAccounts.actAs on it. The identities that call this service (people, services, workloads, third parties) are covered in the [Google Cloud service identities guide](/guides/google-cloud-service-identities/) and the [workload identities guide](/guides/google-cloud-workload-identities/). ## Roles and settings to avoid, with the documented reason The documented cautions concern impersonation and public access. - Service Account User with a powerful service identity: Google's caution says assigning it indirectly grants the role of the service identity to the user. Its example is a service identity with the Editor role, which lets a user act as an Editor. Google recommends least privilege on the service identity itself. - Cloud Run Admin on a project: full control of every Cloud Run resource in it, including policies. - Granting roles/run.invoker to allUsers: Google's Cloud Functions page shows this binding as the way to allow unauthenticated invocation; use it only for intentionally public endpoints. - Basic roles on the project: highly permissive, per Google, and broader than any Cloud Run role. ## A minimal binding to deploy one service Google's documentation lists the roles a deployer account needs, each on a specific resource. | Role | Resource | Permission it brings | | --- | --- | --- | | roles/run.developer | The Cloud Run resource | run.services.create, run.services.update, run.services.get, run.operations.get | | roles/artifactregistry.reader | The Artifact Registry repository holding the image | artifactregistry.repositories.downloadArtifacts | | roles/iam.serviceAccountUser | The Cloud Run service identity | iam.serviceAccounts.actAs | | roles/run.invoker | The one service, for callers | run.routes.invoke, run.instances.invoke, run.jobs.run | ## Resource-level grants and a dedicated identity narrow access Create a dedicated service identity for each service and grant it only what that service calls, so a deploy permission cannot be turned into a broad role. Grant Invoker per service to named callers rather than all authenticated users. Google's IAM Conditions page says only some resource types accept conditions in role bindings. Where Cloud Run bindings do not accept a condition, place a conditional grant at the project or organization level instead, as that page describes. ## Checking what is used The IAM recommender lists excess permissions on principals using 90 days of usage, and it includes service accounts, according to Google's documentation. Check the service identity of each Cloud Run service with it. Then list IAM policies on each service with the Viewer role's view of policies, and remove invoker grants for callers that no longer call. Two follow-up checks are worth scheduling. First, list which service identity each Cloud Run service uses, and confirm that no service shares the default identity with unrelated workloads. Second, list the principals holding Service Account User on each identity, because Google's caution means each of them can act with that identity's roles. Removing a stale Service Account User grant removes a path to the identity without touching the service. ## Related pages - [Least privilege for Cloud Functions](/services/gcp-cloud-functions/) - [Least privilege for Cloud KMS](/services/gcp-cloud-kms/) - [Least privilege for Pub/Sub](/services/gcp-pubsub/) - [All of them](/services/) ## Sources - Cloud Run IAM roles: https://cloud.google.com/run/docs/reference/iam/roles (read 16 Nov 2026) - Cloud Functions access with IAM: https://cloud.google.com/functions/docs/securing/managing-access-iam (read 16 Nov 2026) - IAM Conditions overview: https://cloud.google.com/iam/docs/conditions-overview (read 16 Nov 2026) - Role recommendations overview: https://cloud.google.com/iam/docs/recommender-overview (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/services/gcp-cloud-run/ --- # Least privilege for Cloud Functions means Cloud Functions Admin for few, Invoker per function and no allUsers Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. Cloud Functions separates administering functions from invoking them. Google's documentation, read 16 Nov 2026, says HTTP functions require authentication by default, that the Owner and Cloud Functions Admin roles hold setIamPolicy, and that the Domain Restricted Sharing policy blocks unauthenticated invocation. ## Cloud Functions uses IAM for administration and a separate invoker role for calling Google's documentation says IAM authorizes administrative actions on functions created with the Cloud Functions v2 API, such as creating, updating and deleting. Roles can be granted on the function or the project. By default, entities that need to invoke an HTTP function must present authentication credentials and have the required permissions. Event-driven functions can only be invoked by the event source they subscribe to. To invoke an HTTP function, the documented role is roles/run.invoker, granted on the specific function. The identities that call this service (people, services, workloads, third parties) are covered in the [Google Cloud service identities guide](/guides/google-cloud-service-identities/) and the [workload identities guide](/guides/google-cloud-workload-identities/). ## Roles and flags to avoid, with the documented reason Three choices change who can reach a function. - Owner and Cloud Functions Admin: Google says cloudfunctions.functions.setIamPolicy, needed to control access to a function, is in both roles. Anyone with it can change who invokes the function. - allUsers as invoker: Google shows it as the way to allow unauthenticated invocation, and says roles cannot be granted to allUsers and allAuthenticatedUsers at the project level. - The --allow-unauthenticated flag at deploy: Google says using it modifies IAM permissions, and that later deployments of the same function do not change its status. - Basic roles on the project: highly permissive, per Google's IAM overview. ## A minimal binding for a caller of one HTTP function Google's page shows the invoker grant as a command of the form gcloud run services add-iam-policy-binding FUNCTION_NAME --member= --role=roles/run.invoker, where the member is a specific principal rather than allUsers. | Task | Role | Resource | Note from Google's documentation | | --- | --- | --- | --- | | Call one HTTP function | roles/run.invoker | The function | Callers must also present credentials | | Change who can call or manage a function | roles/cloudfunctions.admin | The function or project | Contains cloudfunctions.functions.setIamPolicy | | Public endpoint (deliberate) | roles/run.invoker to allUsers | The function | Blocked where Domain Restricted Sharing applies | ## Organization policy and per-function grants narrow access Google's documentation says that in a project subject to the Domain Restricted Sharing organization policy, unauthenticated invocation of a function cannot be allowed. That makes the organization policy a guardrail on top of the function's IAM policy. Grant roles on the single function instead of the project. A project-level role covers every function, including ones added later. ## Checking what is used The IAM recommender in Google's documentation compares granted permissions with those used in the last 90 days and suggests removing or replacing roles with excess permissions. Run it for the project and read the service accounts that functions run as. Then read each function's IAM policy and check for allUsers or allAuthenticatedUsers invoker entries that no longer serve a public endpoint. Treat the three rights as three reviews. Administration is reviewed by listing who holds setIamPolicy on the function or project. Invocation is reviewed by listing the invoker entries on each function and removing callers that no longer call. Deployment is reviewed by listing who can act as the function's runtime service account, since deploying a function that runs as a powerful account lets the deployer use that account. ## Related pages - [Least privilege for Cloud KMS](/services/gcp-cloud-kms/) - [Least privilege for Pub/Sub](/services/gcp-pubsub/) - [Least privilege for Amazon S3](/services/aws-s3/) - [All of them](/services/) ## Sources - Cloud Functions access with IAM: https://cloud.google.com/functions/docs/securing/managing-access-iam (read 16 Nov 2026) - IAM roles overview: https://cloud.google.com/iam/docs/roles-overview (read 16 Nov 2026) - Role recommendations overview: https://cloud.google.com/iam/docs/recommender-overview (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/services/gcp-cloud-functions/ --- # Least privilege for Cloud KMS means a key-level role such as cryptoKeyEncrypterDecrypter, never Owner Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. Cloud KMS binds roles at project, key ring and key level. Google's documentation, read 16 Nov 2026, says to grant the most limited permissions to the lowest object in the hierarchy, and that only Owner or Cloud KMS Admin can grant or revoke access to Cloud KMS resources. ## Cloud KMS roles inherit down from project to key ring to key Google's documentation says policy bindings can be set on the project, key ring, key, import job and other Cloud KMS resources. A principal with a role at a higher level inherits it on child resources, so an owner on a project is an owner on all key rings and keys in it. The inverse does not hold: a permission on a key gives nothing on its key ring. Google notes that a user with permissions only on a key ring or key cannot manage key resources in the console without also holding a browser role on the project. Google also publishes a separation of duties page, which pairs with the role list below. The identities that call this service (people, services, workloads, third parties) are covered in the [Google Cloud service identities guide](/guides/google-cloud-service-identities/) and the [workload identities guide](/guides/google-cloud-workload-identities/). ## Roles to avoid, with the documented reason Administration and use of a key should not be held by the same identity. - Owner (roles/owner) on a project: Google says it makes the holder an owner on every key ring and key in the project. - Cloud KMS Admin (roles/cloudkms.admin): Google's guidance is to grant it on the key to manage that key. Google says only principals with Owner or this role can grant or revoke access to Cloud KMS resources. - cryptoKeyEncrypterDecrypter where encryption alone is enough: choose the encrypter role if the principal never decrypts. - Key-ring or project scope for a single workload: it widens the grant to every key in the ring or project. ## Minimal bindings by task Google's least privilege section maps tasks to roles on the key. | Task | Role on the key | | --- | --- | | Encrypt but not decrypt | roles/cloudkms.cryptoKeyEncrypter | | Encrypt and decrypt | roles/cloudkms.cryptoKeyEncrypterDecrypter | | Verify but not sign | roles/cloudkms.publicKeyViewer | | Sign and verify | roles/cloudkms.signerVerifier | | Manage a key | roles/cloudkms.admin | ## Hierarchy position is the narrowing control Google's page says to practice least privilege by granting the most limited set of permissions to the lowest object in the resource hierarchy. For Cloud KMS the lowest object is usually the key itself, which means a policy per key rather than a policy per project. Because a key ring grant covers every key in it, put keys with different users in different key rings. Pair this with the separation of duties page so the identity that administers keys is not the identity that uses them. ## Checking what is used The IAM recommender compares each principal's roles with the permissions it used in the last 90 days, per Google's documentation. Review the output for principals holding roles/cloudkms.admin or the encrypter and decrypter roles at project scope. Then list the IAM policy of each key and key ring and remove principals that are no longer in the data path. Google's IAM overview states that basic roles include thousands of permissions across all services, and a project-level Editor or Owner would carry key permissions with them. When reviewing Cloud KMS, therefore, list basic role holders on the project as well as the bindings on the keys. A key with a tight policy still inherits whatever the project policy grants above it, so the project policy is part of the key's access. ## Related pages - [Least privilege for Pub/Sub](/services/gcp-pubsub/) - [Least privilege for Amazon S3](/services/aws-s3/) - [Least privilege for AWS Lambda](/services/aws-lambda/) - [All of them](/services/) ## Sources - Cloud KMS access control with IAM: https://cloud.google.com/kms/docs/iam (read 16 Nov 2026) - Role recommendations overview: https://cloud.google.com/iam/docs/recommender-overview (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/services/gcp-cloud-kms/ --- # Least privilege for Pub/Sub means Publisher on a topic and Subscriber on a subscription, not Editor Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. Pub/Sub roles can be granted on individual topics and subscriptions. Google's documentation, read 16 Nov 2026, lists Publisher, Subscriber, Editor and Admin, says custom roles help enforce least privilege, and shows cross-project access using the Subscriber role on one subscription instead of a project-wide grant. ## Pub/Sub roles attach to topics, subscriptions, snapshots and schemas Google's documentation says Pub/Sub supports basic, predefined and custom roles, and that custom roles help enforce the principle of least privilege. The predefined roles can be granted on resources as small as a single topic or subscription. Pub/Sub Publisher (roles/pubsub.publisher) publishes messages to a topic. Pub/Sub Subscriber (roles/pubsub.subscriber) consumes messages from a subscription and attaches subscriptions to a topic. Pub/Sub Editor (roles/pubsub.editor) modifies topics and subscriptions and can publish and consume. Pub/Sub Admin (roles/pubsub.admin) has full access to topics and subscriptions. The identities that call this service (people, services, workloads, third parties) are covered in the [Google Cloud service identities guide](/guides/google-cloud-service-identities/) and the [workload identities guide](/guides/google-cloud-workload-identities/). ## Roles to avoid, with the documented reason Editor and Admin combine producing, consuming and changing. - Pub/Sub Editor: Google's description is access to modify topics and subscriptions as well as to publish and consume. A producer needs only publish. - Pub/Sub Admin: full access to topics and subscriptions, including the ability to change who can use them. - Project-level grants: a project-level Publisher covers every topic. Grant on the topic instead. - Public or broad principals: the documentation shows bindings to allUsers and domain principals as examples, which are the shape of grant to avoid for message data. ## A minimal binding for a producer and a consumer Google's documentation describes a cross-project case: to let a service account in project A receive messages from a subscription in project B, grant it Pub/Sub Subscriber on project B or, more narrowly, on the subscription. | Task | Role | Resource | Detail in Google's documentation | | --- | --- | --- | --- | | Publish to a topic | roles/pubsub.publisher | The topic | Includes pubsub.topics.publish | | Receive from a subscription | roles/pubsub.subscriber | The subscription | Consume messages; attach subscriptions to a topic | | Create a subscription on another project's topic | roles/pubsub.editor, or a custom role | Project A and the topic | Needs pubsub.subscriptions.create and pubsub.topics.attachSubscription | ## Custom roles and per-resource grants narrow access The cross-project example needs two permissions, pubsub.subscriptions.create on project A and pubsub.topics.attachSubscription on the topic in project B. The Editor role holds both, but a custom role with only those two is narrower, and Google presents custom roles as the least privilege tool. Google notes that Pub/Sub may fall back on cached IAM policies to keep latency low, so changes take a short time to apply. Plan revocations with that delay in mind. ## Checking what is used The IAM recommender compares a principal's roles with its used permissions over the last 90 days, per Google's documentation. Check principals that hold Editor or Admin and find those that only publish or only consume. Then list the IAM policy of each topic and subscription and remove principals that no longer take part in the flow. Google's documentation also lists the permissions that each API method on topics and subscriptions needs. Use that table to build a custom role for each producer and consumer from the methods they call, rather than starting from Editor and removing permissions. Start with the methods in your code: publish for a producer, and pull or acknowledge for a consumer. ## Related pages - [Least privilege for Amazon S3](/services/aws-s3/) - [Least privilege for AWS Lambda](/services/aws-lambda/) - [Least privilege for Amazon EC2](/services/aws-ec2/) - [All of them](/services/) ## Sources - Pub/Sub access control: https://cloud.google.com/pubsub/docs/access-control (read 16 Nov 2026) - IAM roles overview: https://cloud.google.com/iam/docs/roles-overview (read 16 Nov 2026) - Role recommendations overview: https://cloud.google.com/iam/docs/recommender-overview (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/services/gcp-pubsub/ --- # Routes to least privilege: provider tools, right-sizing products and boundary platforms Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. Routes sets out the different ways a team can carry out a least privilege program in the cloud and what each route asks of the team. Each route is described by who does the work, what it needs as input and where it tends to stop, and the page ranks none of them. Every entry is described by the same fields from the pages read on the date shown, in alphabetical order. No entry carries a number of merit, and the order is the alphabet; what each says about itself is labelled as its own statement. | Name | Route | What it changes | Clouds | Who applies the change | Public pricing | Source | | --- | --- | --- | --- | --- | --- | --- | | [Act Security](/routes/act/) | Boundary platform | Act says its human-identity boundary rightsizes human access and removes standing privileged access, and that an AI boundary contains agents so they cannot reach beyond their intended scope. | Listed on AWS, Azure and GCP marketplaces; Act says, in the SACR report, that the same policy applies across every cloud provider. | The SACR report, which lists Act among the vendors that took part in its research program, says Act has read-only access and delivers changes through Terraform, coding agents, tickets and DevOps workflows, so the customer carries out the removal. Act says changes arrive as pull requests to Infrastructure-as-Code. | none found on 2026-11-16 (the pricing page returns Not Found; a free assessment and a demo booking are offered) | [read 16 Nov 2026](https://act.security/platform?utm_source=leastprivilegeprinciple.com&utm_medium=referral&utm_campaign=directory) | | [Apono](/routes/apono/) | Time-bound access product | Apono says it grants just-in-time access to humans, machines and AI agents and revokes it automatically, and that its access discovery can restrict an over-privileged AWS principal with a permission boundary. | AWS, Azure, GCP, Kubernetes and databases (named on the home page). | Apono says an access flow sets the lifetime, approval is automatic or by approvers, and Apono revokes the access when the lifetime ends. A connector runs in the customer environment. | none found on 2026-11-16 (the pricing page shows a free trial button and no prices) | [read 16 Nov 2026](https://www.apono.io/) | | [AWS Control Tower](/routes/aws-control-tower/) | Provider tool | AWS documentation says preventive controls deny the action and log denied attempts in CloudTrail, and that a preventive control enabled on an organizational unit is enforced on its member accounts. | AWS only (multi-account AWS Organizations environments). | AWS operates the service; the customer enables controls on organizational units, and AWS enforces them. Control Tower builds on AWS Organizations, AWS Service Catalog and IAM Identity Center. | AWS says there is no additional charge for Control Tower; the AWS services it enables are billed (read 2026-11-10). | [read 10 Nov 2026](https://docs.aws.amazon.com/controltower/latest/userguide/pricing.html) | | [AWS IAM Access Analyzer](/routes/aws-iam-access-analyzer/) | Provider tool | AWS says Access Analyzer reports unused roles, access keys, passwords and permissions, recommends a refined policy for unused permissions, and generates policies from access activity. | AWS only. | The console gives quick links to delete unused roles, keys and passwords and the user performs the deletion. For unused permissions it recommends a refined policy and the page read does not say AWS applies it. | AWS lists unused access analysis at $0.20 per IAM role or user analyzed per month and internal access analysis at $9.00 per AWS resource analyzed per month (read 2026-11-10). | [read 10 Nov 2026](https://aws.amazon.com/iam/access-analyzer/pricing/) | | [AWS IAM Identity Center temporary elevated access (TEAM)](/routes/aws-identity-center-tea/) | Provider tool | AWS describes TEAM, an open source solution, as letting users request access to an AWS account for a specific period and removing the elevated access automatically when the period ends. | AWS only (multi-account AWS Organizations environments through IAM Identity Center permission sets). | The customer deploys TEAM in its own AWS environment; approvers review requests and the solution removes access at expiry. Eligibility policies set the account or OU, permission sets, a maximum duration of 1 to 8000 hours, and whether approval is required. | TEAM is open source (MIT-0) with no license fee; the customer pays for the AWS services it runs on (read 2026-11-16). | [read 16 Nov 2026](https://aws-samples.github.io/iam-identity-center-team/docs/overview/cost.html) | | [Britive](/routes/britive/) | Time-bound access product | Britive says privileged access is created when needed and removed automatically when the task ends, and that non-human identities such as pipelines lose permanent credentials and receive access when a job runs. | AWS, Azure, Google Cloud and OCI (named on the home page). | Britive says it provisions and removes access at runtime. The pages read do not state who deploys the agent or connector. | none found on 2026-11-16 (the pricing page returns 404; a demo request is offered) | [read 16 Nov 2026](https://www.britive.com/) | | [ConductorOne (C1.ai)](/routes/conductorone/) | Time-bound access product | ConductorOne says C1 provisions access for a set duration and removes it automatically when the time runs out, with approval policies that can route on identity, role, resource and duration. | Connectors documented for AWS, Azure and Google Cloud Platform. | ConductorOne says no automation has to be built once a maximum request duration is set; C1 performs the grant and the removal. | none found on 2026-11-16 for prices (the pricing page names Platform and Flex models and says exact pricing is scoped with the buyer) | [read 16 Nov 2026](https://www.conductorone.com/pricing/) | | [CrowdStrike Falcon Cloud Security](/routes/crowdstrike/) | Right-sizing product | CrowdStrike says its CIEM capability generates least-privilege access policies. | AWS, Azure, Google Cloud and Kubernetes; OCI is also listed under providers. | CrowdStrike's CIEM explainer says CIEM tools generate remediation recommendations and the page read does not state that Falcon applies the change. | none found on 2026-11-10 for Falcon Cloud Security (the pricing page read lists per-device endpoint plans) | [read 10 Nov 2026](https://www.crowdstrike.com/en-us/platform/cloud-security/cnapp/) | | [CyberArk Secure Cloud Access (Palo Alto Networks)](/routes/cyberark/) | Time-bound access product | CyberArk says Secure Cloud Access applies zero standing privileges: it removes standing permissions and grants entitlements just-in-time. | AWS, Azure and GCP (2025 documentation lists AWS IAM, IAM Identity Center, Azure resource roles, Entra ID roles and Google Cloud roles). | CyberArk's documentation describes requesting temporary access; the pages read do not say who applies each change in the cloud account. | none found on 2026-11-16 (the pricing page redirects to the Idira platform page) | [read 16 Nov 2026](https://docs.cyberark.com/) | | [Entitle (BeyondTrust)](/routes/entitle/) | Time-bound access product | BeyondTrust says Entitle automates just-in-time access and permission management and reduces standing privileges for cloud and SaaS, with self-service access that is revoked after a set duration, ticket resolution or on-call rotation. | AWS, Azure and GCP (documented in the Entitle docs). | Entitle says it grants and revokes access through its integrations after a request and approval. | none found on 2026-11-16 (the pricing page is a request form) | [read 16 Nov 2026](https://www.beyondtrust.com/products/entitle) | | [Google Cloud IAM recommender](/routes/google-iam-recommender/) | Provider tool | Google says the IAM recommender produces role recommendations that replace or remove a role, and that Policy Intelligence adds Policy Analyzer, Policy Simulator and Policy Troubleshooter. | Google Cloud only. | A person reviews each recommendation and decides whether to apply or dismiss it; the IAM recommender does not apply recommendations itself. | Google says role recommendations for basic roles are free; other recommendation types and policy insights need the Premium or Enterprise tier of Security Command Center (read 2026-11-10). | [read 10 Nov 2026](https://cloud.google.com/policy-intelligence/docs/role-recommendations-overview) | | [Google Cloud Privileged Access Manager](/routes/gcp-pam/) | Provider tool | Google's documentation says an entitlement defines requesters, roles, a maximum grant duration and optional approvers, and that PAM revokes the granted roles at the end of the grant duration. | Google Cloud projects, folders and organizations only. | Google operates PAM; the customer creates entitlements. The maximum grant duration ranges from 30 minutes to 168 hours. | none found on 2026-11-16 on the IAM pricing page or the Security Command Center pricing page | [read 16 Nov 2026](https://cloud.google.com/iam/pricing) | | [Microsoft Defender for Cloud permissions management (CIEM)](/routes/microsoft/) | Provider tool | Microsoft says Entra Permissions Management is retired and that CIEM continues in Defender for Cloud as an extension of the Defender CSPM plan; the retired product documented automatic removal of permissions unused for 90 days. | Azure, AWS and GCP. | Defender CSPM is enabled per subscription, AWS account or GCP project; the pages read do not say Defender applies the permission changes itself. | Entra Permissions Management was priced at $125 per resource per year and is no longer for purchase; none found on 2026-11-10 for a Defender CSPM price (the page read shows a placeholder). | [read 10 Nov 2026](https://learn.microsoft.com/en-us/entra/permissions-management/faqs) | | [Microsoft Entra Privileged Identity Management](/routes/entra-pim/) | Provider tool | Microsoft's documentation says PIM separates eligible assignments, which must be activated, from active assignments, so a user can hold a privileged role without the role being in use. | Azure resources (management groups, subscriptions, resource groups, resources), Microsoft Entra roles and PIM for Groups. | The customer configures eligible assignments and activation settings in PIM; Microsoft operates the service. Roles can be assigned to users, groups, service principals or managed identities. | Microsoft says PIM requires Microsoft Entra ID Governance or Microsoft Entra ID P2 licenses; the licensing page read shows no separate PIM price (read 2026-11-16). | [read 16 Nov 2026](https://learn.microsoft.com/en-us/entra/id-governance/licensing-fundamentals) | | [Opal Security](/routes/opal/) | Time-bound access product | Opal says users can request time-bounded access to AWS IAM roles, EC2, EKS, RDS and Identity Center permission sets, with configurable approval options and reviewers. | AWS (IAM roles, Identity Center, EC2, EKS, RDS), GCP and Azure. | Opal says requests are approved by the configured reviewers and Opal grants the access for a bounded time. | none found on 2026-11-16 (the pricing page presents a Request a Demo form) | [read 16 Nov 2026](https://docs.opal.dev/integrations/setting-up-your-aws-organization-in-opal.md) | | [P0 Security](/routes/p0-security/) | Time-bound access product | P0 says it provisions only the privilege required for the task and removes it when the work is done, across users, non-human identities and AI agents. | AWS, Azure, Google Cloud and Kubernetes (named in the docs index). | P0 says it provisions and removes privileges at request time. The pages read do not state who deploys its connector. | none found on 2026-11-16 (the pricing page returns 404; the docs mention a free starter tier sign-up) | [read 16 Nov 2026](https://www.p0.dev/platform) | | [Prisma Cloud by Palo Alto Networks](/routes/prisma/) | Right-sizing product | Palo Alto Networks documents effective permissions analysis for cloud identities in Prisma Cloud and markets Cortex Cloud as the next version. | AWS, Azure, Google Cloud and OCI (Cortex Cloud CIEM page). | Prisma Cloud documentation says the IAM module has no built-in automatic remediation: fixes are CLI commands the user runs, or a custom script the user configures. Automatic remediation of GCP IAM alerts is not supported. | none found on 2026-11-10 (the pricing URL returned 404; the FAQ says credit-based licensing) | [read 10 Nov 2026](https://docs.prismacloud.io/content-collections/administration/configure-iam-security/remediate-alerts-for-iam-security.md) | | [Sonrai Security](/routes/sonrai/) | Boundary platform | Sonrai says its Cloud Permissions Firewall blocks unused privileged permissions through org-level policies, exempts identities that are using a permission, and quarantines dormant identities. | AWS, Azure and Google Cloud. | Sonrai says policies are built from usage data and nothing enforces until the customer deploys them; enforcement uses cloud-native organization-level policies, and a quarantined identity can be reactivated through ChatOps approval. | Sonrai lists a Startup tier at a $15k per year minimum for 10 to 50 accounts at $150 per account per month; 50 or more accounts are custom pricing (read 2026-11-16). | [read 16 Nov 2026](https://sonraisecurity.com/pricing/) | | [StrongDM (Delinea)](/routes/strongdm/) | Time-bound access product | StrongDM says it enforces time-bound, policy-based access with nothing left standing between logins. | Resource types for AWS (CLI, console, instance profile), Azure and GCP (service account, workload identity federation). | StrongDM's AWS documentation lets the administrator choose between just-in-time requests and standing access for particular users or roles, so standing access is a supported option. | none found on 2026-11-16 (the pricing page shows Delinea Platform tiers named Essentials, Standard and Enterprise with a Get Pricing form and no figures) | [read 16 Nov 2026](https://www.strongdm.com/pricing) | | [Sysdig Secure](/routes/sysdig/) | Right-sizing product | Sysdig says it analyzes access patterns, generates access policies that restrict privileges to what users need, and suggests least-privilege versions of existing policies. | Cloud log-based detections for AWS, GCP and Azure (also Okta and GitHub). | Sysdig's CIEM page says it generates policies and suggestions; the page read does not say Sysdig applies them to the account. | none found on 2026-11-10 (the pricing page offers Request a quote; licensing is by number of hosts) | [read 10 Nov 2026](https://www.sysdig.com/pricing-secure) | | [Teleport](/routes/teleport/) | Time-bound access product | Teleport says it provides just-in-time privileges that expire and no standing privileges, across humans, machines and AI agents; Access Requests give elevated privileges for a limited period. | Application-access guides for AWS, Google Cloud and Azure. | In Access Requests a configurable number of approvers approve or deny, and Teleport issues and expires the elevated role. | none found on 2026-11-16 for prices (usage-based on active users and protected resources, with a Get Pricing form) | [read 16 Nov 2026](https://goteleport.com/pricing/) | | [Tenable Cloud Security](/routes/tenable/) | Right-sizing product | Tenable says its CIEM engine identifies excessive entitlements and enforces least-privilege policies, and that just-in-time access replaces standing access with eligibility to request temporary access. | AWS, Azure and GCP; just-in-time access for AWS works through IAM Identity Center. | For just-in-time access in AWS, Tenable creates a permission set, assigns it to the user and deletes it at expiry; in Azure it creates and deletes role assignments. For right-sizing, Tenable refers to automated remediation workflows and the page read does not say who applies the change. | none found on 2026-11-16 as a price list (the pricing page says cost depends on billable cloud resources and points to a quote form) | [read 16 Nov 2026](https://www.tenable.com/cloud-security/pricing) | | [Turbot Guardrails](/routes/turbot/) | Boundary platform | Turbot says Guardrails fixes misconfigurations automatically from a stream of change events, and in Enforce mode applies the fix itself. | AWS, Azure, GCP, OCI, GitHub, Anthropic and OpenAI. | Turbot says in Enforce mode Guardrails applies the fix itself. It runs as hosted SaaS (Cloud) or self-hosted (Enterprise). | Turbot lists Cloud at $0.05 per control per resource per month with no minimum, and Enterprise at $0.10 per control per resource per month with annual control packs from $50,000 (read 2026-11-10). | [read 10 Nov 2026](https://turbot.com/guardrails/pricing) | | [Wiz](/routes/wiz/) | Right-sizing product | Wiz says its CIEM page auto-generates least-privilege policies to help teams remediate. | Wiz says it works across all major clouds. | The CIEM page read does not say Wiz applies the generated policies. Wiz says elsewhere that it generates plans, assigns ownership and executes fixes. | none found on 2026-11-10 as a price (the page offers a custom quote with Wiz One, Wiz Go and A la Carte options) | [read 10 Nov 2026](https://www.wiz.io/solutions/ciem) | ## Related pages - [The least privilege data note: every edition](/data-note/) - [Guides](/guides/) - [Glossary](/glossary/) ## Sources - Act Security: https://act.security/platform (read 16 Nov 2026) - Apono: https://www.apono.io/ (read 16 Nov 2026) - AWS Control Tower: https://docs.aws.amazon.com/controltower/latest/userguide/pricing.html (read 10 Nov 2026) - AWS IAM Access Analyzer: https://aws.amazon.com/iam/access-analyzer/pricing/ (read 10 Nov 2026) - AWS IAM Identity Center temporary elevated access (TEAM): https://aws-samples.github.io/iam-identity-center-team/docs/overview/cost.html (read 16 Nov 2026) - Britive: https://www.britive.com/ (read 16 Nov 2026) - ConductorOne (C1.ai): https://www.conductorone.com/pricing/ (read 16 Nov 2026) - CrowdStrike Falcon Cloud Security: https://www.crowdstrike.com/en-us/platform/cloud-security/cnapp/ (read 10 Nov 2026) - CyberArk Secure Cloud Access (Palo Alto Networks): https://docs.cyberark.com/ (read 16 Nov 2026) - Entitle (BeyondTrust): https://www.beyondtrust.com/products/entitle (read 16 Nov 2026) - Google Cloud IAM recommender: https://cloud.google.com/policy-intelligence/docs/role-recommendations-overview (read 10 Nov 2026) - Google Cloud Privileged Access Manager: https://cloud.google.com/iam/pricing (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/routes/ --- # Act Security as a route to least privilege Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. Act Security appears in the routes directory as a boundary platform. Act says its human-identity boundary rightsizes human access and removes standing privileged access, and that an AI boundary contains agents so they cannot reach beyond their intended scope. Fields read on 16 Nov 2026 from the source linked on this page. - **Official name:** Act Security - **Website:** [act.security](https://act.security/?utm_source=leastprivilegeprinciple.com&utm_medium=referral&utm_campaign=route) - **Route:** Boundary platform - **What it changes:** Act says its human-identity boundary rightsizes human access and removes standing privileged access, and that an AI boundary contains agents so they cannot reach beyond their intended scope. - **Clouds:** Listed on AWS, Azure and GCP marketplaces; Act says, in the SACR report, that the same policy applies across every cloud provider. - **Who applies the change:** The SACR report, which lists Act among the vendors that took part in its research program, says Act has read-only access and delivers changes through Terraform, coding agents, tickets and DevOps workflows, so the customer carries out the removal. Act says changes arrive as pull requests to Infrastructure-as-Code. - **Public pricing:** none found on 2026-11-16 (the pricing page returns Not Found; a free assessment and a demo booking are offered) - **Read:** 16 Nov 2026 ## Other entries on the same route - [Sonrai Security](/routes/sonrai/) - [Turbot Guardrails](/routes/turbot/) ## Where this route fits The [principle](/principle-of-least-privilege/) and the [first 90 days](/first-90-days/) explain where a boundary platform sits in a least privilege program; the [calculator](/) sizes the work it would replace. ## Related pages - [Apono: time-bound access product for least privilege](/routes/apono/) - [AWS Control Tower: provider tool for least privilege](/routes/aws-control-tower/) - [AWS IAM Access Analyzer: provider tool for least privilege](/routes/aws-iam-access-analyzer/) - [All of them](/routes/) ## Sources - Act Security: https://act.security/platform (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/routes/act/ --- # Apono as a route to least privilege Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. Apono appears in the routes directory as a time-bound access product. Apono says it grants just-in-time access to humans, machines and AI agents and revokes it automatically, and that its access discovery can restrict an over-privileged AWS principal with a permission boundary. - **Official name:** Apono - **Website:** [www.apono.io](https://www.apono.io) - **Route:** Time-bound access product - **What it changes:** Apono says it grants just-in-time access to humans, machines and AI agents and revokes it automatically, and that its access discovery can restrict an over-privileged AWS principal with a permission boundary. - **Clouds:** AWS, Azure, GCP, Kubernetes and databases (named on the home page). - **Who applies the change:** Apono says an access flow sets the lifetime, approval is automatic or by approvers, and Apono revokes the access when the lifetime ends. A connector runs in the customer environment. - **Public pricing:** none found on 2026-11-16 (the pricing page shows a free trial button and no prices) - **Read:** 16 Nov 2026 ## Other entries on the same route - [Britive](/routes/britive/) - [ConductorOne (C1.ai)](/routes/conductorone/) - [CyberArk Secure Cloud Access (Palo Alto Networks)](/routes/cyberark/) - [Entitle (BeyondTrust)](/routes/entitle/) - [Opal Security](/routes/opal/) - [P0 Security](/routes/p0-security/) - [StrongDM (Delinea)](/routes/strongdm/) - [Teleport](/routes/teleport/) ## Where this route fits The [principle](/principle-of-least-privilege/) and the [first 90 days](/first-90-days/) explain where a time-bound access product sits in a least privilege program; the [calculator](/) sizes the work it would replace. ## Related pages - [AWS Control Tower: provider tool for least privilege](/routes/aws-control-tower/) - [AWS IAM Access Analyzer: provider tool for least privilege](/routes/aws-iam-access-analyzer/) - [AWS IAM Identity Center temporary elevated access (TEAM): ](/routes/aws-identity-center-tea/) - [All of them](/routes/) ## Sources - Apono: https://www.apono.io/ (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/routes/apono/ --- # AWS Control Tower as a route to least privilege Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. AWS Control Tower appears in the routes directory as a provider tool. AWS documentation says preventive controls deny the action and log denied attempts in CloudTrail, and that a preventive control enabled on an organizational unit is enforced on its member accounts. - **Official name:** AWS Control Tower - **Website:** [docs.aws.amazon.com](https://docs.aws.amazon.com) - **Route:** Provider tool - **What it changes:** AWS documentation says preventive controls deny the action and log denied attempts in CloudTrail, and that a preventive control enabled on an organizational unit is enforced on its member accounts. - **Clouds:** AWS only (multi-account AWS Organizations environments). - **Who applies the change:** AWS operates the service; the customer enables controls on organizational units, and AWS enforces them. Control Tower builds on AWS Organizations, AWS Service Catalog and IAM Identity Center. - **Public pricing:** AWS says there is no additional charge for Control Tower; the AWS services it enables are billed (read 2026-11-10). - **Read:** 10 Nov 2026 ## Other entries on the same route - [AWS IAM Access Analyzer](/routes/aws-iam-access-analyzer/) - [AWS IAM Identity Center temporary elevated access (TEAM)](/routes/aws-identity-center-tea/) - [Google Cloud IAM recommender](/routes/google-iam-recommender/) - [Google Cloud Privileged Access Manager](/routes/gcp-pam/) - [Microsoft Defender for Cloud permissions management (CIEM)](/routes/microsoft/) - [Microsoft Entra Privileged Identity Management](/routes/entra-pim/) ## Where this route fits The [principle](/principle-of-least-privilege/) and the [first 90 days](/first-90-days/) explain where a provider tool sits in a least privilege program; the [calculator](/) sizes the work it would replace. ## Related pages - [AWS IAM Access Analyzer: provider tool for least privilege](/routes/aws-iam-access-analyzer/) - [AWS IAM Identity Center temporary elevated access (TEAM): ](/routes/aws-identity-center-tea/) - [Britive: time-bound access product for least privilege](/routes/britive/) - [All of them](/routes/) ## Sources - AWS Control Tower: https://docs.aws.amazon.com/controltower/latest/userguide/pricing.html (read 10 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/routes/aws-control-tower/ --- # AWS IAM Access Analyzer as a route to least privilege Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. AWS IAM Access Analyzer appears in the routes directory as a provider tool. AWS says Access Analyzer reports unused roles, access keys, passwords and permissions, recommends a refined policy for unused permissions, and generates policies from access activity. Clouds as documented: AWS only. - **Official name:** AWS IAM Access Analyzer - **Website:** [docs.aws.amazon.com](https://docs.aws.amazon.com) - **Route:** Provider tool - **What it changes:** AWS says Access Analyzer reports unused roles, access keys, passwords and permissions, recommends a refined policy for unused permissions, and generates policies from access activity. - **Clouds:** AWS only. - **Who applies the change:** The console gives quick links to delete unused roles, keys and passwords and the user performs the deletion. For unused permissions it recommends a refined policy and the page read does not say AWS applies it. - **Public pricing:** AWS lists unused access analysis at $0.20 per IAM role or user analyzed per month and internal access analysis at $9.00 per AWS resource analyzed per month (read 2026-11-10). - **Read:** 10 Nov 2026 ## Other entries on the same route - [AWS Control Tower](/routes/aws-control-tower/) - [AWS IAM Identity Center temporary elevated access (TEAM)](/routes/aws-identity-center-tea/) - [Google Cloud IAM recommender](/routes/google-iam-recommender/) - [Google Cloud Privileged Access Manager](/routes/gcp-pam/) - [Microsoft Defender for Cloud permissions management (CIEM)](/routes/microsoft/) - [Microsoft Entra Privileged Identity Management](/routes/entra-pim/) ## Where this route fits The [principle](/principle-of-least-privilege/) and the [first 90 days](/first-90-days/) explain where a provider tool sits in a least privilege program; the [calculator](/) sizes the work it would replace. ## Related pages - [AWS IAM Identity Center temporary elevated access (TEAM): ](/routes/aws-identity-center-tea/) - [Britive: time-bound access product for least privilege](/routes/britive/) - [ConductorOne (C1.ai): time-bound access product for least ](/routes/conductorone/) - [All of them](/routes/) ## Sources - AWS IAM Access Analyzer: https://aws.amazon.com/iam/access-analyzer/pricing/ (read 10 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/routes/aws-iam-access-analyzer/ --- # AWS IAM Identity Center temporary elevated access (TEAM) as a route to least privilege Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. AWS IAM Identity Center temporary elevated access (TEAM) appears in the routes directory as a provider tool. AWS describes TEAM, an open source solution, as letting users request access to an AWS account for a specific period and removing the elevated access automatically when the period ends. - **Official name:** AWS IAM Identity Center temporary elevated access (TEAM) - **Website:** [source](https://aws-samples.github.io/iam-identity-center-team/docs/overview/cost.html) - **Route:** Provider tool - **What it changes:** AWS describes TEAM, an open source solution, as letting users request access to an AWS account for a specific period and removing the elevated access automatically when the period ends. - **Clouds:** AWS only (multi-account AWS Organizations environments through IAM Identity Center permission sets). - **Who applies the change:** The customer deploys TEAM in its own AWS environment; approvers review requests and the solution removes access at expiry. Eligibility policies set the account or OU, permission sets, a maximum duration of 1 to 8000 hours, and whether approval is required. - **Public pricing:** TEAM is open source (MIT-0) with no license fee; the customer pays for the AWS services it runs on (read 2026-11-16). - **Read:** 16 Nov 2026 ## Other entries on the same route - [AWS Control Tower](/routes/aws-control-tower/) - [AWS IAM Access Analyzer](/routes/aws-iam-access-analyzer/) - [Google Cloud IAM recommender](/routes/google-iam-recommender/) - [Google Cloud Privileged Access Manager](/routes/gcp-pam/) - [Microsoft Defender for Cloud permissions management (CIEM)](/routes/microsoft/) - [Microsoft Entra Privileged Identity Management](/routes/entra-pim/) ## Where this route fits The [principle](/principle-of-least-privilege/) and the [first 90 days](/first-90-days/) explain where a provider tool sits in a least privilege program; the [calculator](/) sizes the work it would replace. ## Related pages - [Britive: time-bound access product for least privilege](/routes/britive/) - [ConductorOne (C1.ai): time-bound access product for least ](/routes/conductorone/) - [CrowdStrike Falcon Cloud Security: right-sizing product fo](/routes/crowdstrike/) - [All of them](/routes/) ## Sources - AWS IAM Identity Center temporary elevated access (TEAM): https://aws-samples.github.io/iam-identity-center-team/docs/overview/cost.html (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/routes/aws-identity-center-tea/ --- # Britive as a route to least privilege Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. Britive appears in the routes directory as a time-bound access product. Britive says privileged access is created when needed and removed automatically when the task ends, and that non-human identities such as pipelines lose permanent credentials and receive access when a job runs. - **Official name:** Britive - **Website:** [www.britive.com](https://www.britive.com) - **Route:** Time-bound access product - **What it changes:** Britive says privileged access is created when needed and removed automatically when the task ends, and that non-human identities such as pipelines lose permanent credentials and receive access when a job runs. - **Clouds:** AWS, Azure, Google Cloud and OCI (named on the home page). - **Who applies the change:** Britive says it provisions and removes access at runtime. The pages read do not state who deploys the agent or connector. - **Public pricing:** none found on 2026-11-16 (the pricing page returns 404; a demo request is offered) - **Read:** 16 Nov 2026 ## Other entries on the same route - [Apono](/routes/apono/) - [ConductorOne (C1.ai)](/routes/conductorone/) - [CyberArk Secure Cloud Access (Palo Alto Networks)](/routes/cyberark/) - [Entitle (BeyondTrust)](/routes/entitle/) - [Opal Security](/routes/opal/) - [P0 Security](/routes/p0-security/) - [StrongDM (Delinea)](/routes/strongdm/) - [Teleport](/routes/teleport/) ## Where this route fits The [principle](/principle-of-least-privilege/) and the [first 90 days](/first-90-days/) explain where a time-bound access product sits in a least privilege program; the [calculator](/) sizes the work it would replace. ## Related pages - [ConductorOne (C1.ai): time-bound access product for least ](/routes/conductorone/) - [CrowdStrike Falcon Cloud Security: right-sizing product fo](/routes/crowdstrike/) - [CyberArk Secure Cloud Access (Palo Alto Networks): time-bo](/routes/cyberark/) - [All of them](/routes/) ## Sources - Britive: https://www.britive.com/ (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/routes/britive/ --- # ConductorOne (C1.ai) as a route to least privilege Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. ConductorOne (C1.ai) appears in the routes directory as a time-bound access product. ConductorOne says C1 provisions access for a set duration and removes it automatically when the time runs out, with approval policies that can route on identity, role, resource and duration. - **Official name:** ConductorOne (C1.ai) - **Website:** [source](https://www.conductorone.com/pricing/) - **Route:** Time-bound access product - **What it changes:** ConductorOne says C1 provisions access for a set duration and removes it automatically when the time runs out, with approval policies that can route on identity, role, resource and duration. - **Clouds:** Connectors documented for AWS, Azure and Google Cloud Platform. - **Who applies the change:** ConductorOne says no automation has to be built once a maximum request duration is set; C1 performs the grant and the removal. - **Public pricing:** none found on 2026-11-16 for prices (the pricing page names Platform and Flex models and says exact pricing is scoped with the buyer) - **Read:** 16 Nov 2026 ## Other entries on the same route - [Apono](/routes/apono/) - [Britive](/routes/britive/) - [CyberArk Secure Cloud Access (Palo Alto Networks)](/routes/cyberark/) - [Entitle (BeyondTrust)](/routes/entitle/) - [Opal Security](/routes/opal/) - [P0 Security](/routes/p0-security/) - [StrongDM (Delinea)](/routes/strongdm/) - [Teleport](/routes/teleport/) ## Where this route fits The [principle](/principle-of-least-privilege/) and the [first 90 days](/first-90-days/) explain where a time-bound access product sits in a least privilege program; the [calculator](/) sizes the work it would replace. ## Related pages - [CrowdStrike Falcon Cloud Security: right-sizing product fo](/routes/crowdstrike/) - [CyberArk Secure Cloud Access (Palo Alto Networks): time-bo](/routes/cyberark/) - [Entitle (BeyondTrust): time-bound access product for least](/routes/entitle/) - [All of them](/routes/) ## Sources - ConductorOne (C1.ai): https://www.conductorone.com/pricing/ (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/routes/conductorone/ --- # CrowdStrike Falcon Cloud Security as a route to least privilege Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. CrowdStrike Falcon Cloud Security appears in the routes directory as a right-sizing product. CrowdStrike says its CIEM capability generates least-privilege access policies. Clouds as documented: AWS, Azure, Google Cloud and Kubernetes; OCI is also listed under providers. Fields read on 10 Nov 2026 from the source linked on this page. - **Official name:** CrowdStrike Falcon Cloud Security - **Website:** [www.crowdstrike.com](https://www.crowdstrike.com) - **Route:** Right-sizing product - **What it changes:** CrowdStrike says its CIEM capability generates least-privilege access policies. - **Clouds:** AWS, Azure, Google Cloud and Kubernetes; OCI is also listed under providers. - **Who applies the change:** CrowdStrike's CIEM explainer says CIEM tools generate remediation recommendations and the page read does not state that Falcon applies the change. - **Public pricing:** none found on 2026-11-10 for Falcon Cloud Security (the pricing page read lists per-device endpoint plans) - **Read:** 10 Nov 2026 ## Other entries on the same route - [Prisma Cloud by Palo Alto Networks](/routes/prisma/) - [Sysdig Secure](/routes/sysdig/) - [Tenable Cloud Security](/routes/tenable/) - [Wiz](/routes/wiz/) ## Where this route fits The [principle](/principle-of-least-privilege/) and the [first 90 days](/first-90-days/) explain where a right-sizing product sits in a least privilege program; the [calculator](/) sizes the work it would replace. ## Related pages - [CyberArk Secure Cloud Access (Palo Alto Networks): time-bo](/routes/cyberark/) - [Entitle (BeyondTrust): time-bound access product for least](/routes/entitle/) - [Google Cloud IAM recommender: provider tool for least priv](/routes/google-iam-recommender/) - [All of them](/routes/) ## Sources - CrowdStrike Falcon Cloud Security: https://www.crowdstrike.com/en-us/platform/cloud-security/cnapp/ (read 10 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/routes/crowdstrike/ --- # CyberArk Secure Cloud Access (Palo Alto Networks) as a route to least privilege Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. CyberArk Secure Cloud Access (Palo Alto Networks) appears in the routes directory as a time-bound access product. CyberArk says Secure Cloud Access applies zero standing privileges: it removes standing permissions and grants entitlements just-in-time. Clouds as documented: AWS, Azure and GCP (2025 documentation lists AWS IAM, IAM Identity Center, Azure resource roles, Entra ID roles and Google Cloud roles). - **Official name:** CyberArk Secure Cloud Access (Palo Alto Networks) - **Website:** [source](https://docs.cyberark.com/) - **Route:** Time-bound access product - **What it changes:** CyberArk says Secure Cloud Access applies zero standing privileges: it removes standing permissions and grants entitlements just-in-time. - **Clouds:** AWS, Azure and GCP (2025 documentation lists AWS IAM, IAM Identity Center, Azure resource roles, Entra ID roles and Google Cloud roles). - **Who applies the change:** CyberArk's documentation describes requesting temporary access; the pages read do not say who applies each change in the cloud account. - **Public pricing:** none found on 2026-11-16 (the pricing page redirects to the Idira platform page) - **Read:** 16 Nov 2026 ## Other entries on the same route - [Apono](/routes/apono/) - [Britive](/routes/britive/) - [ConductorOne (C1.ai)](/routes/conductorone/) - [Entitle (BeyondTrust)](/routes/entitle/) - [Opal Security](/routes/opal/) - [P0 Security](/routes/p0-security/) - [StrongDM (Delinea)](/routes/strongdm/) - [Teleport](/routes/teleport/) ## Where this route fits The [principle](/principle-of-least-privilege/) and the [first 90 days](/first-90-days/) explain where a time-bound access product sits in a least privilege program; the [calculator](/) sizes the work it would replace. ## Related pages - [Entitle (BeyondTrust): time-bound access product for least](/routes/entitle/) - [Google Cloud IAM recommender: provider tool for least priv](/routes/google-iam-recommender/) - [Google Cloud Privileged Access Manager: provider tool for ](/routes/gcp-pam/) - [All of them](/routes/) ## Sources - CyberArk Secure Cloud Access (Palo Alto Networks): https://docs.cyberark.com/ (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/routes/cyberark/ --- # Entitle (BeyondTrust) as a route to least privilege Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. Entitle (BeyondTrust) appears in the routes directory as a time-bound access product. BeyondTrust says Entitle automates just-in-time access and permission management and reduces standing privileges for cloud and SaaS, with self-service access that is revoked after a set duration, ticket resolution or on-call rotation. - **Official name:** Entitle (BeyondTrust) - **Website:** [www.beyondtrust.com](https://www.beyondtrust.com) - **Route:** Time-bound access product - **What it changes:** BeyondTrust says Entitle automates just-in-time access and permission management and reduces standing privileges for cloud and SaaS, with self-service access that is revoked after a set duration, ticket resolution or on-call rotation. - **Clouds:** AWS, Azure and GCP (documented in the Entitle docs). - **Who applies the change:** Entitle says it grants and revokes access through its integrations after a request and approval. - **Public pricing:** none found on 2026-11-16 (the pricing page is a request form) - **Read:** 16 Nov 2026 ## Other entries on the same route - [Apono](/routes/apono/) - [Britive](/routes/britive/) - [ConductorOne (C1.ai)](/routes/conductorone/) - [CyberArk Secure Cloud Access (Palo Alto Networks)](/routes/cyberark/) - [Opal Security](/routes/opal/) - [P0 Security](/routes/p0-security/) - [StrongDM (Delinea)](/routes/strongdm/) - [Teleport](/routes/teleport/) ## Where this route fits The [principle](/principle-of-least-privilege/) and the [first 90 days](/first-90-days/) explain where a time-bound access product sits in a least privilege program; the [calculator](/) sizes the work it would replace. ## Related pages - [Google Cloud IAM recommender: provider tool for least priv](/routes/google-iam-recommender/) - [Google Cloud Privileged Access Manager: provider tool for ](/routes/gcp-pam/) - [Microsoft Defender for Cloud permissions management (CIEM)](/routes/microsoft/) - [All of them](/routes/) ## Sources - Entitle (BeyondTrust): https://www.beyondtrust.com/products/entitle (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/routes/entitle/ --- # Google Cloud IAM recommender as a route to least privilege Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. Google Cloud IAM recommender appears in the routes directory as a provider tool. Google says the IAM recommender produces role recommendations that replace or remove a role, and that Policy Intelligence adds Policy Analyzer, Policy Simulator and Policy Troubleshooter. Clouds as documented: Google Cloud only. - **Official name:** Google Cloud IAM recommender - **Website:** [cloud.google.com](https://cloud.google.com) - **Route:** Provider tool - **What it changes:** Google says the IAM recommender produces role recommendations that replace or remove a role, and that Policy Intelligence adds Policy Analyzer, Policy Simulator and Policy Troubleshooter. - **Clouds:** Google Cloud only. - **Who applies the change:** A person reviews each recommendation and decides whether to apply or dismiss it; the IAM recommender does not apply recommendations itself. - **Public pricing:** Google says role recommendations for basic roles are free; other recommendation types and policy insights need the Premium or Enterprise tier of Security Command Center (read 2026-11-10). - **Read:** 10 Nov 2026 ## Other entries on the same route - [AWS Control Tower](/routes/aws-control-tower/) - [AWS IAM Access Analyzer](/routes/aws-iam-access-analyzer/) - [AWS IAM Identity Center temporary elevated access (TEAM)](/routes/aws-identity-center-tea/) - [Google Cloud Privileged Access Manager](/routes/gcp-pam/) - [Microsoft Defender for Cloud permissions management (CIEM)](/routes/microsoft/) - [Microsoft Entra Privileged Identity Management](/routes/entra-pim/) ## Where this route fits The [principle](/principle-of-least-privilege/) and the [first 90 days](/first-90-days/) explain where a provider tool sits in a least privilege program; the [calculator](/) sizes the work it would replace. ## Related pages - [Google Cloud Privileged Access Manager: provider tool for ](/routes/gcp-pam/) - [Microsoft Defender for Cloud permissions management (CIEM)](/routes/microsoft/) - [Microsoft Entra Privileged Identity Management: provider t](/routes/entra-pim/) - [All of them](/routes/) ## Sources - Google Cloud IAM recommender: https://cloud.google.com/policy-intelligence/docs/role-recommendations-overview (read 10 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/routes/google-iam-recommender/ --- # Google Cloud Privileged Access Manager as a route to least privilege Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. Google Cloud Privileged Access Manager appears in the routes directory as a provider tool. Google's documentation says an entitlement defines requesters, roles, a maximum grant duration and optional approvers, and that PAM revokes the granted roles at the end of the grant duration. - **Official name:** Google Cloud Privileged Access Manager - **Website:** [source](https://cloud.google.com/iam/pricing) - **Route:** Provider tool - **What it changes:** Google's documentation says an entitlement defines requesters, roles, a maximum grant duration and optional approvers, and that PAM revokes the granted roles at the end of the grant duration. - **Clouds:** Google Cloud projects, folders and organizations only. - **Who applies the change:** Google operates PAM; the customer creates entitlements. The maximum grant duration ranges from 30 minutes to 168 hours. - **Public pricing:** none found on 2026-11-16 on the IAM pricing page or the Security Command Center pricing page - **Read:** 16 Nov 2026 ## Other entries on the same route - [AWS Control Tower](/routes/aws-control-tower/) - [AWS IAM Access Analyzer](/routes/aws-iam-access-analyzer/) - [AWS IAM Identity Center temporary elevated access (TEAM)](/routes/aws-identity-center-tea/) - [Google Cloud IAM recommender](/routes/google-iam-recommender/) - [Microsoft Defender for Cloud permissions management (CIEM)](/routes/microsoft/) - [Microsoft Entra Privileged Identity Management](/routes/entra-pim/) ## Where this route fits The [principle](/principle-of-least-privilege/) and the [first 90 days](/first-90-days/) explain where a provider tool sits in a least privilege program; the [calculator](/) sizes the work it would replace. ## Related pages - [Microsoft Defender for Cloud permissions management (CIEM)](/routes/microsoft/) - [Microsoft Entra Privileged Identity Management: provider t](/routes/entra-pim/) - [Opal Security: time-bound access product for least privile](/routes/opal/) - [All of them](/routes/) ## Sources - Google Cloud Privileged Access Manager: https://cloud.google.com/iam/pricing (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/routes/gcp-pam/ --- # Microsoft Defender for Cloud permissions management (CIEM) as a route to least privilege Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. Microsoft Defender for Cloud permissions management (CIEM) appears in the routes directory as a provider tool. Microsoft says Entra Permissions Management is retired and that CIEM continues in Defender for Cloud as an extension of the Defender CSPM plan; the retired product documented automatic removal of permissions unused for 90 days. - **Official name:** Microsoft Defender for Cloud permissions management (CIEM) - **Website:** [learn.microsoft.com](https://learn.microsoft.com) - **Route:** Provider tool - **What it changes:** Microsoft says Entra Permissions Management is retired and that CIEM continues in Defender for Cloud as an extension of the Defender CSPM plan; the retired product documented automatic removal of permissions unused for 90 days. - **Clouds:** Azure, AWS and GCP. - **Who applies the change:** Defender CSPM is enabled per subscription, AWS account or GCP project; the pages read do not say Defender applies the permission changes itself. - **Public pricing:** Entra Permissions Management was priced at $125 per resource per year and is no longer for purchase; none found on 2026-11-10 for a Defender CSPM price (the page read shows a placeholder). - **Read:** 10 Nov 2026 ## Other entries on the same route - [AWS Control Tower](/routes/aws-control-tower/) - [AWS IAM Access Analyzer](/routes/aws-iam-access-analyzer/) - [AWS IAM Identity Center temporary elevated access (TEAM)](/routes/aws-identity-center-tea/) - [Google Cloud IAM recommender](/routes/google-iam-recommender/) - [Google Cloud Privileged Access Manager](/routes/gcp-pam/) - [Microsoft Entra Privileged Identity Management](/routes/entra-pim/) ## Where this route fits The [principle](/principle-of-least-privilege/) and the [first 90 days](/first-90-days/) explain where a provider tool sits in a least privilege program; the [calculator](/) sizes the work it would replace. ## Related pages - [Microsoft Entra Privileged Identity Management: provider t](/routes/entra-pim/) - [Opal Security: time-bound access product for least privile](/routes/opal/) - [P0 Security: time-bound access product for least privilege](/routes/p0-security/) - [All of them](/routes/) ## Sources - Microsoft Defender for Cloud permissions management (CIEM): https://learn.microsoft.com/en-us/entra/permissions-management/faqs (read 10 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/routes/microsoft/ --- # Microsoft Entra Privileged Identity Management as a route to least privilege Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. Microsoft Entra Privileged Identity Management appears in the routes directory as a provider tool. Microsoft's documentation says PIM separates eligible assignments, which must be activated, from active assignments, so a user can hold a privileged role without the role being in use. - **Official name:** Microsoft Entra Privileged Identity Management - **Website:** [source](https://learn.microsoft.com/en-us/entra/id-governance/licensing-fundamentals) - **Route:** Provider tool - **What it changes:** Microsoft's documentation says PIM separates eligible assignments, which must be activated, from active assignments, so a user can hold a privileged role without the role being in use. - **Clouds:** Azure resources (management groups, subscriptions, resource groups, resources), Microsoft Entra roles and PIM for Groups. - **Who applies the change:** The customer configures eligible assignments and activation settings in PIM; Microsoft operates the service. Roles can be assigned to users, groups, service principals or managed identities. - **Public pricing:** Microsoft says PIM requires Microsoft Entra ID Governance or Microsoft Entra ID P2 licenses; the licensing page read shows no separate PIM price (read 2026-11-16). - **Read:** 16 Nov 2026 ## Other entries on the same route - [AWS Control Tower](/routes/aws-control-tower/) - [AWS IAM Access Analyzer](/routes/aws-iam-access-analyzer/) - [AWS IAM Identity Center temporary elevated access (TEAM)](/routes/aws-identity-center-tea/) - [Google Cloud IAM recommender](/routes/google-iam-recommender/) - [Google Cloud Privileged Access Manager](/routes/gcp-pam/) - [Microsoft Defender for Cloud permissions management (CIEM)](/routes/microsoft/) ## Where this route fits The [principle](/principle-of-least-privilege/) and the [first 90 days](/first-90-days/) explain where a provider tool sits in a least privilege program; the [calculator](/) sizes the work it would replace. ## Related pages - [Opal Security: time-bound access product for least privile](/routes/opal/) - [P0 Security: time-bound access product for least privilege](/routes/p0-security/) - [Prisma Cloud by Palo Alto Networks: right-sizing product f](/routes/prisma/) - [All of them](/routes/) ## Sources - Microsoft Entra Privileged Identity Management: https://learn.microsoft.com/en-us/entra/id-governance/licensing-fundamentals (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/routes/entra-pim/ --- # Opal Security as a route to least privilege Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. Opal Security appears in the routes directory as a time-bound access product. Opal says users can request time-bounded access to AWS IAM roles, EC2, EKS, RDS and Identity Center permission sets, with configurable approval options and reviewers. Clouds as documented: AWS (IAM roles, Identity Center, EC2, EKS, RDS), GCP and Azure. - **Official name:** Opal Security - **Website:** [source](https://docs.opal.dev/integrations/setting-up-your-aws-organization-in-opal.md) - **Route:** Time-bound access product - **What it changes:** Opal says users can request time-bounded access to AWS IAM roles, EC2, EKS, RDS and Identity Center permission sets, with configurable approval options and reviewers. - **Clouds:** AWS (IAM roles, Identity Center, EC2, EKS, RDS), GCP and Azure. - **Who applies the change:** Opal says requests are approved by the configured reviewers and Opal grants the access for a bounded time. - **Public pricing:** none found on 2026-11-16 (the pricing page presents a Request a Demo form) - **Read:** 16 Nov 2026 ## Other entries on the same route - [Apono](/routes/apono/) - [Britive](/routes/britive/) - [ConductorOne (C1.ai)](/routes/conductorone/) - [CyberArk Secure Cloud Access (Palo Alto Networks)](/routes/cyberark/) - [Entitle (BeyondTrust)](/routes/entitle/) - [P0 Security](/routes/p0-security/) - [StrongDM (Delinea)](/routes/strongdm/) - [Teleport](/routes/teleport/) ## Where this route fits The [principle](/principle-of-least-privilege/) and the [first 90 days](/first-90-days/) explain where a time-bound access product sits in a least privilege program; the [calculator](/) sizes the work it would replace. ## Related pages - [P0 Security: time-bound access product for least privilege](/routes/p0-security/) - [Prisma Cloud by Palo Alto Networks: right-sizing product f](/routes/prisma/) - [Sonrai Security: boundary platform for least privilege](/routes/sonrai/) - [All of them](/routes/) ## Sources - Opal Security: https://docs.opal.dev/integrations/setting-up-your-aws-organization-in-opal.md (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/routes/opal/ --- # P0 Security as a route to least privilege Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. P0 Security appears in the routes directory as a time-bound access product. P0 says it provisions only the privilege required for the task and removes it when the work is done, across users, non-human identities and AI agents. Clouds as documented: AWS, Azure, Google Cloud and Kubernetes (named in the docs index). - **Official name:** P0 Security - **Website:** [source](https://www.p0.dev/platform) - **Route:** Time-bound access product - **What it changes:** P0 says it provisions only the privilege required for the task and removes it when the work is done, across users, non-human identities and AI agents. - **Clouds:** AWS, Azure, Google Cloud and Kubernetes (named in the docs index). - **Who applies the change:** P0 says it provisions and removes privileges at request time. The pages read do not state who deploys its connector. - **Public pricing:** none found on 2026-11-16 (the pricing page returns 404; the docs mention a free starter tier sign-up) - **Read:** 16 Nov 2026 ## Other entries on the same route - [Apono](/routes/apono/) - [Britive](/routes/britive/) - [ConductorOne (C1.ai)](/routes/conductorone/) - [CyberArk Secure Cloud Access (Palo Alto Networks)](/routes/cyberark/) - [Entitle (BeyondTrust)](/routes/entitle/) - [Opal Security](/routes/opal/) - [StrongDM (Delinea)](/routes/strongdm/) - [Teleport](/routes/teleport/) ## Where this route fits The [principle](/principle-of-least-privilege/) and the [first 90 days](/first-90-days/) explain where a time-bound access product sits in a least privilege program; the [calculator](/) sizes the work it would replace. ## Related pages - [Prisma Cloud by Palo Alto Networks: right-sizing product f](/routes/prisma/) - [Sonrai Security: boundary platform for least privilege](/routes/sonrai/) - [StrongDM (Delinea): time-bound access product for least pr](/routes/strongdm/) - [All of them](/routes/) ## Sources - P0 Security: https://www.p0.dev/platform (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/routes/p0-security/ --- # Prisma Cloud by Palo Alto Networks as a route to least privilege Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. Prisma Cloud by Palo Alto Networks appears in the routes directory as a right-sizing product. Palo Alto Networks documents effective permissions analysis for cloud identities in Prisma Cloud and markets Cortex Cloud as the next version. Clouds as documented: AWS, Azure, Google Cloud and OCI (Cortex Cloud CIEM page). - **Official name:** Prisma Cloud by Palo Alto Networks - **Website:** [www.paloaltonetworks.com](https://www.paloaltonetworks.com) - **Route:** Right-sizing product - **What it changes:** Palo Alto Networks documents effective permissions analysis for cloud identities in Prisma Cloud and markets Cortex Cloud as the next version. - **Clouds:** AWS, Azure, Google Cloud and OCI (Cortex Cloud CIEM page). - **Who applies the change:** Prisma Cloud documentation says the IAM module has no built-in automatic remediation: fixes are CLI commands the user runs, or a custom script the user configures. Automatic remediation of GCP IAM alerts is not supported. - **Public pricing:** none found on 2026-11-10 (the pricing URL returned 404; the FAQ says credit-based licensing) - **Read:** 10 Nov 2026 ## Other entries on the same route - [CrowdStrike Falcon Cloud Security](/routes/crowdstrike/) - [Sysdig Secure](/routes/sysdig/) - [Tenable Cloud Security](/routes/tenable/) - [Wiz](/routes/wiz/) ## Where this route fits The [principle](/principle-of-least-privilege/) and the [first 90 days](/first-90-days/) explain where a right-sizing product sits in a least privilege program; the [calculator](/) sizes the work it would replace. ## Related pages - [Sonrai Security: boundary platform for least privilege](/routes/sonrai/) - [StrongDM (Delinea): time-bound access product for least pr](/routes/strongdm/) - [Sysdig Secure: right-sizing product for least privilege](/routes/sysdig/) - [All of them](/routes/) ## Sources - Prisma Cloud by Palo Alto Networks: https://docs.prismacloud.io/content-collections/administration/configure-iam-security/remediate-alerts-for-iam-security.md (read 10 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/routes/prisma/ --- # Sonrai Security as a route to least privilege Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. Sonrai Security appears in the routes directory as a boundary platform. Sonrai says its Cloud Permissions Firewall blocks unused privileged permissions through org-level policies, exempts identities that are using a permission, and quarantines dormant identities. Clouds as documented: AWS, Azure and Google Cloud. - **Official name:** Sonrai Security - **Website:** [sonraisecurity.com](https://sonraisecurity.com) - **Route:** Boundary platform - **What it changes:** Sonrai says its Cloud Permissions Firewall blocks unused privileged permissions through org-level policies, exempts identities that are using a permission, and quarantines dormant identities. - **Clouds:** AWS, Azure and Google Cloud. - **Who applies the change:** Sonrai says policies are built from usage data and nothing enforces until the customer deploys them; enforcement uses cloud-native organization-level policies, and a quarantined identity can be reactivated through ChatOps approval. - **Public pricing:** Sonrai lists a Startup tier at a $15k per year minimum for 10 to 50 accounts at $150 per account per month; 50 or more accounts are custom pricing (read 2026-11-16). - **Read:** 16 Nov 2026 ## Other entries on the same route - [Act Security](/routes/act/) - [Turbot Guardrails](/routes/turbot/) ## Where this route fits The [principle](/principle-of-least-privilege/) and the [first 90 days](/first-90-days/) explain where a boundary platform sits in a least privilege program; the [calculator](/) sizes the work it would replace. ## Related pages - [StrongDM (Delinea): time-bound access product for least pr](/routes/strongdm/) - [Sysdig Secure: right-sizing product for least privilege](/routes/sysdig/) - [Teleport: time-bound access product for least privilege](/routes/teleport/) - [All of them](/routes/) ## Sources - Sonrai Security: https://sonraisecurity.com/pricing/ (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/routes/sonrai/ --- # StrongDM (Delinea) as a route to least privilege Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. StrongDM (Delinea) appears in the routes directory as a time-bound access product. StrongDM says it enforces time-bound, policy-based access with nothing left standing between logins. Clouds as documented: Resource types for AWS (CLI, console, instance profile), Azure and GCP (service account, workload identity federation). - **Official name:** StrongDM (Delinea) - **Website:** [source](https://www.strongdm.com/pricing) - **Route:** Time-bound access product - **What it changes:** StrongDM says it enforces time-bound, policy-based access with nothing left standing between logins. - **Clouds:** Resource types for AWS (CLI, console, instance profile), Azure and GCP (service account, workload identity federation). - **Who applies the change:** StrongDM's AWS documentation lets the administrator choose between just-in-time requests and standing access for particular users or roles, so standing access is a supported option. - **Public pricing:** none found on 2026-11-16 (the pricing page shows Delinea Platform tiers named Essentials, Standard and Enterprise with a Get Pricing form and no figures) - **Read:** 16 Nov 2026 ## Other entries on the same route - [Apono](/routes/apono/) - [Britive](/routes/britive/) - [ConductorOne (C1.ai)](/routes/conductorone/) - [CyberArk Secure Cloud Access (Palo Alto Networks)](/routes/cyberark/) - [Entitle (BeyondTrust)](/routes/entitle/) - [Opal Security](/routes/opal/) - [P0 Security](/routes/p0-security/) - [Teleport](/routes/teleport/) ## Where this route fits The [principle](/principle-of-least-privilege/) and the [first 90 days](/first-90-days/) explain where a time-bound access product sits in a least privilege program; the [calculator](/) sizes the work it would replace. ## Related pages - [Sysdig Secure: right-sizing product for least privilege](/routes/sysdig/) - [Teleport: time-bound access product for least privilege](/routes/teleport/) - [Tenable Cloud Security: right-sizing product for least pri](/routes/tenable/) - [All of them](/routes/) ## Sources - StrongDM (Delinea): https://www.strongdm.com/pricing (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/routes/strongdm/ --- # Sysdig Secure as a route to least privilege Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. Sysdig Secure appears in the routes directory as a right-sizing product. Sysdig says it analyzes access patterns, generates access policies that restrict privileges to what users need, and suggests least-privilege versions of existing policies. Clouds as documented: Cloud log-based detections for AWS, GCP and Azure (also Okta and GitHub). - **Official name:** Sysdig Secure - **Website:** [www.sysdig.com](https://www.sysdig.com) - **Route:** Right-sizing product - **What it changes:** Sysdig says it analyzes access patterns, generates access policies that restrict privileges to what users need, and suggests least-privilege versions of existing policies. - **Clouds:** Cloud log-based detections for AWS, GCP and Azure (also Okta and GitHub). - **Who applies the change:** Sysdig's CIEM page says it generates policies and suggestions; the page read does not say Sysdig applies them to the account. - **Public pricing:** none found on 2026-11-10 (the pricing page offers Request a quote; licensing is by number of hosts) - **Read:** 10 Nov 2026 ## Other entries on the same route - [CrowdStrike Falcon Cloud Security](/routes/crowdstrike/) - [Prisma Cloud by Palo Alto Networks](/routes/prisma/) - [Tenable Cloud Security](/routes/tenable/) - [Wiz](/routes/wiz/) ## Where this route fits The [principle](/principle-of-least-privilege/) and the [first 90 days](/first-90-days/) explain where a right-sizing product sits in a least privilege program; the [calculator](/) sizes the work it would replace. ## Related pages - [Teleport: time-bound access product for least privilege](/routes/teleport/) - [Tenable Cloud Security: right-sizing product for least pri](/routes/tenable/) - [Turbot Guardrails: boundary platform for least privilege](/routes/turbot/) - [All of them](/routes/) ## Sources - Sysdig Secure: https://www.sysdig.com/pricing-secure (read 10 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/routes/sysdig/ --- # Teleport as a route to least privilege Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. Teleport appears in the routes directory as a time-bound access product. Teleport says it provides just-in-time privileges that expire and no standing privileges, across humans, machines and AI agents; Access Requests give elevated privileges for a limited period. Clouds as documented: Application-access guides for AWS, Google Cloud and Azure. - **Official name:** Teleport - **Website:** [source](https://goteleport.com/pricing/) - **Route:** Time-bound access product - **What it changes:** Teleport says it provides just-in-time privileges that expire and no standing privileges, across humans, machines and AI agents; Access Requests give elevated privileges for a limited period. - **Clouds:** Application-access guides for AWS, Google Cloud and Azure. - **Who applies the change:** In Access Requests a configurable number of approvers approve or deny, and Teleport issues and expires the elevated role. - **Public pricing:** none found on 2026-11-16 for prices (usage-based on active users and protected resources, with a Get Pricing form) - **Read:** 16 Nov 2026 ## Other entries on the same route - [Apono](/routes/apono/) - [Britive](/routes/britive/) - [ConductorOne (C1.ai)](/routes/conductorone/) - [CyberArk Secure Cloud Access (Palo Alto Networks)](/routes/cyberark/) - [Entitle (BeyondTrust)](/routes/entitle/) - [Opal Security](/routes/opal/) - [P0 Security](/routes/p0-security/) - [StrongDM (Delinea)](/routes/strongdm/) ## Where this route fits The [principle](/principle-of-least-privilege/) and the [first 90 days](/first-90-days/) explain where a time-bound access product sits in a least privilege program; the [calculator](/) sizes the work it would replace. ## Related pages - [Tenable Cloud Security: right-sizing product for least pri](/routes/tenable/) - [Turbot Guardrails: boundary platform for least privilege](/routes/turbot/) - [Wiz: right-sizing product for least privilege](/routes/wiz/) - [All of them](/routes/) ## Sources - Teleport: https://goteleport.com/pricing/ (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/routes/teleport/ --- # Tenable Cloud Security as a route to least privilege Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. Tenable Cloud Security appears in the routes directory as a right-sizing product. Tenable says its CIEM engine identifies excessive entitlements and enforces least-privilege policies, and that just-in-time access replaces standing access with eligibility to request temporary access. Clouds as documented: AWS, Azure and GCP; just-in-time access for AWS works through IAM Identity Center. - **Official name:** Tenable Cloud Security - **Website:** [www.tenable.com](https://www.tenable.com) - **Route:** Right-sizing product - **What it changes:** Tenable says its CIEM engine identifies excessive entitlements and enforces least-privilege policies, and that just-in-time access replaces standing access with eligibility to request temporary access. - **Clouds:** AWS, Azure and GCP; just-in-time access for AWS works through IAM Identity Center. - **Who applies the change:** For just-in-time access in AWS, Tenable creates a permission set, assigns it to the user and deletes it at expiry; in Azure it creates and deletes role assignments. For right-sizing, Tenable refers to automated remediation workflows and the page read does not say who applies the change. - **Public pricing:** none found on 2026-11-16 as a price list (the pricing page says cost depends on billable cloud resources and points to a quote form) - **Read:** 16 Nov 2026 ## Other entries on the same route - [CrowdStrike Falcon Cloud Security](/routes/crowdstrike/) - [Prisma Cloud by Palo Alto Networks](/routes/prisma/) - [Sysdig Secure](/routes/sysdig/) - [Wiz](/routes/wiz/) ## Where this route fits The [principle](/principle-of-least-privilege/) and the [first 90 days](/first-90-days/) explain where a right-sizing product sits in a least privilege program; the [calculator](/) sizes the work it would replace. ## Related pages - [Turbot Guardrails: boundary platform for least privilege](/routes/turbot/) - [Wiz: right-sizing product for least privilege](/routes/wiz/) - [Act Security: boundary platform for least privilege](/routes/act/) - [All of them](/routes/) ## Sources - Tenable Cloud Security: https://www.tenable.com/cloud-security/pricing (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/routes/tenable/ --- # Turbot Guardrails as a route to least privilege Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. Turbot Guardrails appears in the routes directory as a boundary platform. Turbot says Guardrails fixes misconfigurations automatically from a stream of change events, and in Enforce mode applies the fix itself. Clouds as documented: AWS, Azure, GCP, OCI, GitHub, Anthropic and OpenAI. - **Official name:** Turbot Guardrails - **Website:** [turbot.com](https://turbot.com) - **Route:** Boundary platform - **What it changes:** Turbot says Guardrails fixes misconfigurations automatically from a stream of change events, and in Enforce mode applies the fix itself. - **Clouds:** AWS, Azure, GCP, OCI, GitHub, Anthropic and OpenAI. - **Who applies the change:** Turbot says in Enforce mode Guardrails applies the fix itself. It runs as hosted SaaS (Cloud) or self-hosted (Enterprise). - **Public pricing:** Turbot lists Cloud at $0.05 per control per resource per month with no minimum, and Enterprise at $0.10 per control per resource per month with annual control packs from $50,000 (read 2026-11-10). - **Read:** 10 Nov 2026 ## Other entries on the same route - [Act Security](/routes/act/) - [Sonrai Security](/routes/sonrai/) ## Where this route fits The [principle](/principle-of-least-privilege/) and the [first 90 days](/first-90-days/) explain where a boundary platform sits in a least privilege program; the [calculator](/) sizes the work it would replace. ## Related pages - [Wiz: right-sizing product for least privilege](/routes/wiz/) - [Act Security: boundary platform for least privilege](/routes/act/) - [Apono: time-bound access product for least privilege](/routes/apono/) - [All of them](/routes/) ## Sources - Turbot Guardrails: https://turbot.com/guardrails/pricing (read 10 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/routes/turbot/ --- # Wiz as a route to least privilege Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. Wiz appears in the routes directory as a right-sizing product. Wiz says its CIEM page auto-generates least-privilege policies to help teams remediate. Clouds as documented: Wiz says it works across all major clouds. Fields read on 10 Nov 2026 from the source linked on this page. - **Official name:** Wiz - **Website:** [www.wiz.io](https://www.wiz.io) - **Route:** Right-sizing product - **What it changes:** Wiz says its CIEM page auto-generates least-privilege policies to help teams remediate. - **Clouds:** Wiz says it works across all major clouds. - **Who applies the change:** The CIEM page read does not say Wiz applies the generated policies. Wiz says elsewhere that it generates plans, assigns ownership and executes fixes. - **Public pricing:** none found on 2026-11-10 as a price (the page offers a custom quote with Wiz One, Wiz Go and A la Carte options) - **Read:** 10 Nov 2026 ## Other entries on the same route - [CrowdStrike Falcon Cloud Security](/routes/crowdstrike/) - [Prisma Cloud by Palo Alto Networks](/routes/prisma/) - [Sysdig Secure](/routes/sysdig/) - [Tenable Cloud Security](/routes/tenable/) ## Where this route fits The [principle](/principle-of-least-privilege/) and the [first 90 days](/first-90-days/) explain where a right-sizing product sits in a least privilege program; the [calculator](/) sizes the work it would replace. ## Related pages - [Act Security: boundary platform for least privilege](/routes/act/) - [Apono: time-bound access product for least privilege](/routes/apono/) - [AWS Control Tower: provider tool for least privilege](/routes/aws-control-tower/) - [All of them](/routes/) ## Sources - Wiz: https://www.wiz.io/solutions/ciem (read 10 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/routes/wiz/ --- # The monthly least privilege data note Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. The calculator's data note explains where its inputs come from and what they mean. Counts such as identities with unused permissions come from provider reports read on 16 Nov 2026, and the note states each report's window and limits, so a number entered today is comparable with one entered later. Each edition is assembled from the dated items this site records: a change to a provider's least privilege tooling, a new or revised publication, or a documented change to how a permission works. Every item is read from its primary source (AWS, Microsoft, Google Cloud or NIST) and carries the date the source gives it. An edition adds no commentary beyond one plain sentence per item, and nothing appears that the source does not say. There is no mailing list yet. The feed below carries every item as it is added, and each edition stays at its own address so it can be cited. - [Data note, November 2026](/data-note/2026-11/): 1 dated items from primary sources - [Data note, October 2026](/data-note/2026-10/): 3 dated items from primary sources - [Data note, September 2026](/data-note/2026-09/): 5 dated items from primary sources - [Data note, August 2026](/data-note/2026-08/): 3 dated items from primary sources ## Related pages - [Guides](/guides/) - [Glossary](/glossary/) - [News](/news/) ## Sources - https://aws.amazon.com/blogs/security/introducing-aws-account-access-manager-iam-role-flexibility-with-centralized-workforce-access/ (read 16 Nov 2026) - https://cloud.google.com/iam/docs/release-notes (read 16 Nov 2026) - https://cloud.google.com/iam/docs/release-notes (read 16 Nov 2026) - https://aws.amazon.com/about-aws/whats-new/2026/10/aws-identity-store-network-controls/ (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/data-note/ --- # Least privilege data note, November 2026 Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. The November 2026 edition collects 1 dated least privilege changes read from primary sources: AWS introduces account access manager for assigning IAM roles to workforce users. Each item links to its full page and its source, read on 16 Nov 2026. ## AWS introduces account access manager for assigning IAM roles to workforce users 5 Nov 2026. The AWS Security Blog described account access manager on 5 November 2026. AWS says it assigns existing IAM roles from individual accounts to users and groups in IAM Identity Center, keeps role tags, paths and trust policy conditions, and records the individual user in each role session for audit. [Read the item](/news/aws-account-access-manager/). ## Related pages - [Least privilege data note, October 2026](/data-note/2026-10/) - [Least privilege data note, September 2026](/data-note/2026-09/) - [Least privilege data note, August 2026](/data-note/2026-08/) - [All of them](/news/) ## Sources - https://aws.amazon.com/blogs/security/introducing-aws-account-access-manager-iam-role-flexibility-with-centralized-workforce-access/ (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/data-note/2026-11/ --- # Least privilege data note, October 2026 Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. The October 2026 edition collects 3 dated least privilege changes read from primary sources: Google Cloud basic roles Admin, Writer and Reader are generally available; Google Cloud can revoke Workforce Identity Federation sessions in general availability; IAM Identity Center adds network access controls for Identity Store. Each item links to its full page and its source, read on 16 Nov. ## Google Cloud basic roles Admin, Writer and Reader are generally available 29 Oct 2026. Google Cloud IAM release notes dated 29 October 2026 say the Admin, Writer and Reader basic roles are generally available. Google recommends them over the legacy Owner, Editor and Viewer roles when a basic role must be granted, because the newer roles support IAM Conditions. [Read the item](/news/google-basic-roles-admin-writer-reader-ga/). ## Google Cloud can revoke Workforce Identity Federation sessions in general availability 6 Oct 2026. Google Cloud IAM release notes dated 6 October 2026 say revoking active sessions and short-lived credentials for Workforce Identity Federation users is generally available across all clients. Google says administrators can revoke workforce user sessions with the gcloud CLI or the REST API. [Read the item](/news/google-workforce-session-revocation-ga/). ## IAM Identity Center adds network access controls for Identity Store 5 Oct 2026. AWS announced on 5 October 2026 that IAM Identity Center supports network access controls for Identity Store. AWS says administrators can limit the Identity Store API and the SCIM API to allowed VPC endpoints, source VPCs or IP ranges, with different rules per API. The controls are optional and off by default. [Read the item](/news/identity-center-identity-store-network-controls/). ## Related pages - [Least privilege data note, September 2026](/data-note/2026-09/) - [Least privilege data note, August 2026](/data-note/2026-08/) - [AWS introduces account access manager for assigning IAM roles to workforce users](/news/aws-account-access-manager/) - [All of them](/news/) ## Sources - https://cloud.google.com/iam/docs/release-notes (read 16 Nov 2026) - https://cloud.google.com/iam/docs/release-notes (read 16 Nov 2026) - https://aws.amazon.com/about-aws/whats-new/2026/10/aws-identity-store-network-controls/ (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/data-note/2026-10/ --- # Least privilege data note, September 2026 Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. The September 2026 edition collects 5 dated least privilege changes read from primary sources: Policy Troubleshooter now explains access for agent identities; Amazon ECS extends CPU and memory condition keys to RunTask and StartTask; Google Cloud IAM MCP server reaches general availability for custom roles and deny policies. Each item links to its full page and its source, read on. ## Policy Troubleshooter now explains access for agent identities 15 Sep 2026. Google Cloud Policy Intelligence release notes dated 15 September 2026 say Policy Troubleshooter supports agent identities. Google says it can troubleshoot IAM allow policies, deny policies and principal access boundary policies from an agent principal identifier or from an error ID in an access denial event. [Read the item](/news/google-policy-troubleshooter-agent-identities/). ## Amazon ECS extends CPU and memory condition keys to RunTask and StartTask 10 Sep 2026. AWS announced on 10 September 2026 that the ecs:task-cpu and ecs:task-memory IAM condition keys now apply to the RunTask and StartTask APIs. AWS says administrators can enforce the same CPU and memory limits however a task is launched, at no additional cost in all Regions where ECS is available. [Read the item](/news/ecs-cpu-memory-condition-keys-runtask/). ## Google Cloud IAM MCP server reaches general availability for custom roles and deny policies 10 Sep 2026. Google Cloud IAM release notes dated 10 September 2026 say the IAM remote MCP server is generally available. Google says AI applications can connect to it to inspect and manage custom roles and deny policies across resources, two of the main tools for narrowing access below a predefined role. [Read the item](/news/google-iam-mcp-server-ga/). ## Google Cloud Policy Assist API returns IAM role suggestions in Preview 9 Sep 2026. Google Cloud release notes dated 9 September 2026 say the Policy Assist REST API and a Policy Assist remote MCP server are in Preview. Google says Policy Assist returns IAM role suggestions for individual principals with AI assistance, so teams can request a narrower predefined role by code instead of in the console. [Read the item](/news/google-policy-assist-api-preview/). ## Google Cloud Policy Analyzer MCP server is generally available 1 Sep 2026. Google Cloud release notes dated 1 September 2026 say the Policy Analyzer MCP server is generally available. Google says it lets agents and AI applications analyze and audit IAM configurations, which extends Policy Analyzer, the tool for finding which principals can access which resources, to programmatic callers. [Read the item](/news/google-policy-analyzer-mcp-server-ga/). ## Related pages - [Least privilege data note, August 2026](/data-note/2026-08/) - [AWS introduces account access manager for assigning IAM roles to workforce users](/news/aws-account-access-manager/) - [Google Cloud basic roles Admin, Writer and Reader are generally available](/news/google-basic-roles-admin-writer-reader-ga/) - [All of them](/news/) ## Sources - https://cloud.google.com/policy-intelligence/docs/release-notes (read 16 Nov 2026) - https://aws.amazon.com/about-aws/whats-new/2026/09/amazon-ecs-expands-condition-key-support/ (read 16 Nov 2026) - https://cloud.google.com/iam/docs/release-notes (read 16 Nov 2026) - https://cloud.google.com/policy-intelligence/docs/release-notes (read 16 Nov 2026) - https://cloud.google.com/policy-intelligence/docs/release-notes (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/data-note/2026-09/ --- # Least privilege data note, August 2026 Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. The August 2026 edition collects 3 dated least privilege changes read from primary sources: Lambda functions support full IAM resource-based policies; IAM raises the default managed policies per role from 10 to 20; IAM Policy Autopilot generates policies from Terraform plan files. Each item links to its full page and its source, read on 16 Nov 2026. ## Lambda functions support full IAM resource-based policies 25 Aug 2026. AWS announced on 25 August 2026 that Lambda functions now support full IAM resource-based policies. AWS says one policy document can cover multiple principals and actions and use the full range of IAM condition keys, such as source IP or principal tag, at no additional charge in all commercial Regions. [Read the item](/news/lambda-full-iam-resource-based-policies/). ## IAM raises the default managed policies per role from 10 to 20 19 Aug 2026. AWS announced on 19 August 2026 that the default quota for managed policies attached to one IAM role rose from 10 to 20. AWS says teams can now split permissions into purpose-specific policies without a quota request, and can request up to 25 through Service Quotas if they need more. [Read the item](/news/iam-20-managed-policies-per-role/). ## IAM Policy Autopilot generates policies from Terraform plan files 18 Aug 2026. AWS announced on 18 August 2026 that IAM Policy Autopilot, an open source tool, can now generate baseline IAM policies from a Terraform plan file. AWS says the policies cover the create, read, update and delete calls for the resources in the plan and name resource ARNs instead of wildcards where possible. [Read the item](/news/iam-policy-autopilot-terraform-plan/). ## Related pages - [AWS introduces account access manager for assigning IAM roles to workforce users](/news/aws-account-access-manager/) - [Google Cloud basic roles Admin, Writer and Reader are generally available](/news/google-basic-roles-admin-writer-reader-ga/) - [Google Cloud can revoke Workforce Identity Federation sessions in general availability](/news/google-workforce-session-revocation-ga/) - [All of them](/news/) ## Sources - https://aws.amazon.com/about-aws/whats-new/2026/08/aws-lambda-full-iam-resource-based-policies/ (read 16 Nov 2026) - https://aws.amazon.com/about-aws/whats-new/2026/08/aws-iam-quota-increase/ (read 16 Nov 2026) - https://aws.amazon.com/about-aws/whats-new/2026/08/iam-policy-autopilot-now-supports-terraform-plan-files (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/data-note/2026-08/ --- # Guides Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. Guides to least privilege in AWS, Azure and Google Cloud, with one set for each identity type: human, service, workload and third-party. Each guide names the provider mechanisms, quotes the documentation and gives the date it was read, so a team can compare the three clouds on one page. - [Least privilege for human users in AWS](/guides/aws-human-identities/): Permission sets, IAM users, Access Analyzer and last accessed data for people. - [Least privilege for service identities in AWS](/guides/aws-service-identities/): Service-linked roles, service roles and long-term keys for non-human callers. - [Least privilege for workload identities in AWS](/guides/aws-workload-identities/): Lambda execution roles, IRSA, EKS Pod Identity and Roles Anywhere. - [Least privilege for third-party identities in AWS](/guides/aws-third-party-identities/): Cross-account roles for vendors and partners, and how to cut them. - [Least privilege for human users in Azure](/guides/azure-human-identities/): Entra users and groups, RBAC assignments, PIM, access reviews. - [Least privilege for service identities in Azure](/guides/azure-service-identities/): App registrations, service principals and their credentials. - [Least privilege for workload identities in Azure](/guides/azure-workload-identities/): Managed identities, workload identity federation and their role assignments. - [Least privilege for third-party identities in Azure](/guides/azure-third-party-identities/): Azure Lighthouse, multitenant apps and B2B guests. - [Least privilege for human users in Google Cloud](/guides/google-cloud-human-identities/): Users, groups, IAM roles, the IAM recommender and Policy Intelligence. - [Least privilege for service identities in Google Cloud](/guides/google-cloud-service-identities/): Service accounts, service account keys and service agents. - [Least privilege for workload identities in Google Cloud](/guides/google-cloud-workload-identities/): Attached service accounts, Workload Identity Federation and Kubernetes. - [Least privilege for third-party identities in Google Cloud](/guides/google-cloud-third-party-identities/): Cross-project and cross-organization access and federation for partners. ## Related pages - [Glossary](/glossary/) - [News](/news/) - [Least privilege for 20 cloud services](/services/) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/guides/ --- # Least privilege for human users in AWS starts with permission sets, not IAM users Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. Human users in AWS get permissions through IAM Identity Center permission sets, federated roles or IAM users. AWS's IAM guide, read 16 Nov 2026, recommends temporary credentials, Access Analyzer unused access findings and generated policies. This guide gives the steps, the policies to avoid and what to measure. ## Human users reach AWS through permission sets, federated roles or IAM users AWS's IAM best practices page defines human identities as the administrators, developers, operators and consumers of your applications, and tells you to require them to use federation with an identity provider and temporary credentials. For central management it recommends AWS IAM Identity Center. In IAM Identity Center a permission set is a template that holds one or more IAM policies. When you assign a permission set to a user or group, AWS says IAM Identity Center creates a matching IAM role in each target account and attaches the policies to it. Changing the permission set updates those roles. The other routes are IAM users with console passwords or access keys, and the account root user. AWS lists narrow cases where an IAM user is still needed, such as third-party clients that do not support IAM Identity Center. Where an IAM user or root user exists, AWS says to require multi-factor authentication. Each route is a place where privilege accumulates. [Privilege creep](/privilege-creep/) in AWS usually shows up as a permission set that began as AdministratorAccess and was never replaced, or as an IAM user whose keys nobody has used for a year. ## AWS provides three tools that compare granted access with used access IAM Access Analyzer unused access analyzers monitor all IAM roles and users in an organization or account. AWS documents findings for unused roles, unused IAM user access keys and passwords, and unused service-level and action-level permissions within a usage window you set. AWS states that external access findings are free and that unused access findings are charged. IAM last accessed information reports the services and actions an identity last used. AWS documents a tracking period of at least 400 days for service information, with action tracking that began on different dates by service (Amazon S3 on 12 April 2020, EC2, IAM and Lambda on 7 April 2021, other services on 23 May 2023). Recent activity appears in the console within four hours. IAM Access Analyzer policy generation reads AWS CloudTrail events for a date range of up to 90 days and produces a policy template that holds the actions and services the user or role used. You review it, add resources and conditions, and attach it in place of the broad policy. AWS's permission set guidance describes the same sequence: run Access Analyzer against principals that use AWS managed policies, learn what they use, then write or generate a narrower policy. The wider method is in [the first 90 days](/first-90-days/) and [the least privilege policy template](/least-privilege-policy-template/). ## Steps to bring human access in an AWS organization to least privilege - List every route people use: permission sets, federated roles, IAM users, root. Record which accounts each permission set is assigned to. - Move people from IAM users to IAM Identity Center or another federated route, so credentials are temporary. - Give each administrator a second, narrower permission set. AWS's own example assigns PowerUserAccess for development work and keeps AdministratorAccess for administration, and users choose the narrower role in the access portal. - Create an unused access analyzer for the organization and work the findings by oldest unused first. - For each remaining broad permission set, generate a policy from CloudTrail activity (up to 90 days), edit it, test it in a non-production account, then replace the broad policy. - Disable or delete IAM user access keys and passwords that the analyzer reports as unused, using access last used data to confirm. - Set a review date for every permission set and repeat the analyzer review on that date. ## Policies and credentials to avoid for human users | Policy or credential | Documented reason (AWS, read 16 Nov 2026) | Use instead | | --- | --- | --- | | AdministratorAccess | AWS describes it as granting all actions for all AWS services on all resources in the account. | A job-function or custom permission set; a separate administrator permission set used only for administration. | | PowerUserAccess as a permanent default | AWS says it allows all actions for all services except IAM, AWS Organizations and Account Management, so it is a step down from administrator rather than a least-privilege set. | A permission set generated from the user's CloudTrail activity. | | IAM user access keys for people | AWS says IAM users with access keys are an account security risk and that keys in the shared credentials file are stored in plaintext. | IAM Identity Center sign-in with temporary credentials. | | Root user access keys | AWS says not to use root credentials to create access keys. | Root only for tasks that require it, with MFA. | | Unmodified AWS managed policies | AWS says managed policies may not grant least privilege for your use case because they serve all customers. | Customer managed policies scoped to the task. | ## Measure granted permissions against used permissions per principal For each permission set and IAM user, count the services granted and the services last accessed in the tracking window, and record the ratio per principal. Use the unused access analyzer finding count as a second series: unused roles, unused keys and passwords, unused permissions. Also count principals holding AdministratorAccess and the number of accounts each is assigned to. These counts come from the assignments themselves, so they need no usage data and can be taken today. The [Access Debt Calculator](/) on this site turns granted and used counts into a single figure, and [the principle of least privilege](/principle-of-least-privilege/) explains why the unit is the permission rather than the person. Related guides: [Azure human identities](/guides/azure-human-identities/) and [Google Cloud human identities](/guides/google-cloud-human-identities/). ## Related pages - [Least privilege for service identities in AWS](/guides/aws-service-identities/) - [Least privilege for workload identities in AWS](/guides/aws-workload-identities/) - [Least privilege for third-party identities in AWS](/guides/aws-third-party-identities/) - [All of them](/guides/) ## Sources - AWS IAM User Guide: Security best practices in IAM: https://docs.aws.amazon.com/IAM/latest/UserGuide/best-practices.html (read 16 Nov 2026) - AWS IAM Identity Center User Guide: Permission sets: https://docs.aws.amazon.com/singlesignon/latest/userguide/permissionsetsconcept.html (read 16 Nov 2026) - AWS IAM User Guide: Using IAM Access Analyzer: https://docs.aws.amazon.com/IAM/latest/UserGuide/what-is-access-analyzer.html (read 16 Nov 2026) - AWS IAM User Guide: IAM Access Analyzer findings: https://docs.aws.amazon.com/IAM/latest/UserGuide/access-analyzer-findings.html (read 16 Nov 2026) - AWS IAM User Guide: Refine permissions using last accessed information: https://docs.aws.amazon.com/IAM/latest/UserGuide/access_policies_last-accessed.html (read 16 Nov 2026) - AWS IAM User Guide: IAM Access Analyzer policy generation: https://docs.aws.amazon.com/IAM/latest/UserGuide/access-analyzer-policy-generation.html (read 16 Nov 2026) - AWS IAM User Guide: AWS managed policies for job functions: https://docs.aws.amazon.com/IAM/latest/UserGuide/access_policies_job-functions.html (read 16 Nov 2026) - AWS IAM User Guide: Manage access keys for IAM users: https://docs.aws.amazon.com/IAM/latest/UserGuide/id_credentials_access-keys.html (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/guides/aws-human-identities/ --- # Least privilege for service identities in AWS means scoping roles that services assume and retiring access keys Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. Service identities in AWS are the roles and IAM users that services, integrations and third-party tools use to act in an account. AWS's documentation, read 16 Nov 2026, covers service-linked roles, service roles and access keys. This guide lists the controls, the steps and the measures. ## Service identities are service-linked roles, service roles and IAM users with keys This guide uses service identity for a non-human identity that an AWS service or an outside integration uses, as distinct from code running on your own compute, which the [workload identities guide](/guides/aws-workload-identities/) covers. A service-linked role is a role predefined by an AWS service. AWS's documentation says the linked service defines its permissions, an IAM administrator can view but not edit them, and the permissions policy cannot be attached to any other IAM entity. Your control is whether the service may create the role, and monitoring what it does. A service role is a role a service assumes on your behalf. AWS says an IAM administrator can create, modify and delete it, so its permissions are yours to scope. Bedrock Agents and Lambda use roles of this kind. IAM users with long-term access keys complete the set. AWS's best practices page names workloads that cannot use roles, third-party clients that do not support IAM Identity Center and some service-specific credentials as the cases where they remain. ## Access Analyzer and last accessed data show which service roles are idle The IAM Access Analyzer unused access analyzer reports unused roles, unused access keys and unused permissions for roles and users in the usage window you choose. AWS says it reviews last accessed information for all roles in the organization to do this. For keys, AWS tells you to use access last used information to update and remove them safely, and says you can keep at most two access keys per IAM user. For roles you own, policy generation reads CloudTrail for up to 90 days and returns a template of the services and actions used. AWS notes that the service role used by Access Analyzer must have access to your CloudTrail trail. That makes policy generation a practical way to cut a service role that was written broadly during a project. Service control policies and resource control policies add a ceiling across accounts. AWS states that they grant nothing by themselves, so they limit service roles without replacing the scoping work. The ceiling idea is covered under [what least privilege access means](/least-privilege-access/). ## Steps to bring service identities to least privilege - Inventory roles by who assumes them: AWS services, third-party vendors, CI/CD systems, other accounts. Mark service-linked roles separately because their permissions are fixed. - List IAM users that act as service accounts. For each, record the key age, the access last used date and the owner. - Replace keys with roles wherever the caller can assume one: IAM Roles Anywhere for certificate holders, AssumeRoleWithWebIdentity for OIDC tokens, AssumeRoleWithSAML for SAML (all listed in AWS's best practices). - Delete keys and roles that the unused access analyzer reports as unused, after a notice period to the owner. - Generate a policy for each service role you own, review the action list, add resource ARNs and conditions, and attach it. - Where developers may create their own roles, require a permissions boundary so delegated roles cannot exceed the maximum you set. - Review again on a fixed date. Roles drift back to broad grants when a new feature needs one more action. ## Permissions and credentials to avoid on service identities | Policy or credential | Documented reason (AWS, read 16 Nov 2026) | Use instead | | --- | --- | --- | | AdministratorAccess on a service role | AWS describes it as all actions on all resources, which turns any failure of the service into account-wide access. | A policy generated from the role's CloudTrail activity. | | IAM user access keys for an integration that can assume a role | AWS recommends temporary credentials and says IAM users with access keys are an account security risk. | A role with Roles Anywhere, web identity or SAML. | | Keys saved in the shared credentials file | AWS says such keys are stored in plaintext. | A role session from a credential provider. | | Keys that are never rotated or reviewed | AWS says to update keys when needed and to use access last used information to remove them safely. | A scheduled review driven by unused access findings. | | No permissions boundary on delegated roles | AWS says a boundary sets the maximum permissions a delegated identity-based policy can grant. | A permissions boundary on every role that developers create. | ## Measure idle roles, idle keys and unused actions Track four counts per account: service roles with no activity in the window, IAM user keys with no activity in the window, service roles with AdministratorAccess, and for each service role the actions granted against the actions last accessed. AWS's action-level data covers management actions, so record that data events are not part of the comparison. A falling count of idle roles and keys shows clean-up is happening. A flat count of service roles with administrator policies shows the work is stuck at the hardest step. See [how privilege creep builds](/privilege-creep/) and [the routes to least privilege](/routes/) for what to do when the manual loop stops scaling. Related guides: [Azure service identities](/guides/azure-service-identities/) and [Google Cloud service identities](/guides/google-cloud-service-identities/). ## Related pages - [Least privilege for workload identities in AWS](/guides/aws-workload-identities/) - [Least privilege for third-party identities in AWS](/guides/aws-third-party-identities/) - [Least privilege for human users in Azure](/guides/azure-human-identities/) - [All of them](/guides/) ## Sources - AWS IAM User Guide: Security best practices in IAM: https://docs.aws.amazon.com/IAM/latest/UserGuide/best-practices.html (read 16 Nov 2026) - AWS IAM User Guide: Service-linked roles: https://docs.aws.amazon.com/IAM/latest/UserGuide/id_roles_create-service-linked-role.html (read 16 Nov 2026) - AWS IAM User Guide: Using IAM Access Analyzer: https://docs.aws.amazon.com/IAM/latest/UserGuide/what-is-access-analyzer.html (read 16 Nov 2026) - AWS IAM User Guide: IAM Access Analyzer findings: https://docs.aws.amazon.com/IAM/latest/UserGuide/access-analyzer-findings.html (read 16 Nov 2026) - AWS IAM User Guide: Refine permissions using last accessed information: https://docs.aws.amazon.com/IAM/latest/UserGuide/access_policies_last-accessed.html (read 16 Nov 2026) - AWS IAM User Guide: IAM Access Analyzer policy generation: https://docs.aws.amazon.com/IAM/latest/UserGuide/access-analyzer-policy-generation.html (read 16 Nov 2026) - AWS IAM User Guide: Manage access keys for IAM users: https://docs.aws.amazon.com/IAM/latest/UserGuide/id_credentials_access-keys.html (read 16 Nov 2026) - AWS IAM User Guide: AWS managed policies for job functions: https://docs.aws.amazon.com/IAM/latest/UserGuide/access_policies_job-functions.html (read 16 Nov 2026) - AWS IAM User Guide: IAM roles: https://docs.aws.amazon.com/IAM/latest/UserGuide/id_roles.html (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/guides/aws-service-identities/ --- # Least privilege for workload identities in AWS means one narrow role per workload, delivered as temporary credentials Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. Workload identities in AWS are the IAM roles that Lambda functions, EC2 instances and Kubernetes pods use. AWS documentation read on 16 Nov 2026 covers Lambda execution roles, IRSA and EKS Pod Identity. This guide gives the tools, the steps, the roles to avoid and the measures. ## Workloads get permissions from roles that AWS delivers as temporary credentials AWS's IAM best practices page tells you to require workloads to use temporary credentials with IAM roles. On compute services such as Amazon EC2 and Lambda, AWS delivers the role's credentials to the compute resource and AWS SDKs find them without any key being distributed. A Lambda function runs under an execution role. AWS's Lambda guide says Lambda assumes the role when the function is invoked, that the console creates a default role with the AWSLambdaBasicExecutionRole managed policy (permission to write logs to CloudWatch Logs), and that the trust policy must name lambda.amazonaws.com. For Kubernetes on Amazon EKS there are two mechanisms. IAM roles for service accounts (IRSA) uses an OIDC provider for each cluster and the AssumeRoleWithWebIdentity call. EKS Pod Identity maps a role to a Kubernetes service account in a namespace, and the role's trust policy names the principal pods.eks.amazonaws.com. AWS describes Pod Identity as simpler than IRSA because it does not use OIDC identity providers. Workloads outside AWS can still use temporary credentials through IAM Roles Anywhere, AssumeRoleWithSAML, AssumeRoleWithWebIdentity and several service-specific integrations, according to the same best practices page. ## Access Analyzer policy generation is AWS's route to a narrower execution role AWS's Lambda guide says a role created during development often grants more than needed and that before production you should adjust it to the required permissions. It points to IAM Access Analyzer, which reviews CloudTrail logs for the date range you specify and produces a policy template containing only what the function used. The generated template can be turned into a managed policy and attached to the role. AWS states the date range can be up to 90 days, so a job that runs quarterly will be missing from a 90-day window. Test generated policies before production, as AWS advises. The unused access analyzer reports unused roles and unused permissions for existing roles, and IAM last accessed information shows the services and actions each role last used. Both rely on the same CloudTrail and last accessed data, so they agree on what is idle. For pods, the scoping happens at the service account. AWS says you can scope IAM permissions to a service account so only pods using it receive them, and that when access to the instance metadata service is restricted a pod can retrieve credentials only for its own role. AWS also warns that containers are not a security boundary and that pods with hostNetwork set to true always have metadata service access. ## Steps to bring workload roles to least privilege - Make one role per workload, not per team. Name it for the workload so ownership is visible in the console and in CloudTrail. - Remove IAM user access keys from workloads that run on AWS compute and attach a role instead. - On EKS, restrict access to the instance metadata service for pods and associate a separate role with each service account through Pod Identity or IRSA, so pods do not fall back to the node role. - Generate a policy from CloudTrail activity for each execution or task role that currently carries a managed or wildcard policy. - Replace wildcard actions with the generated action list, then add resource ARNs and conditions. AWS's best practices page describes conditions such as requiring requests to use TLS or to come through a specific service. - Where one account hosts many workloads, add a service control policy as a ceiling. AWS says SCPs grant no permissions on their own. - Rerun policy generation after each release that adds an integration, and delete roles the unused access analyzer flags. ## Roles and settings to avoid on workload identities | Role or setting | Documented reason (AWS, read 16 Nov 2026) | Use instead | | --- | --- | --- | | AdministratorAccess on an execution role | AWS describes it as all actions on all resources, and says development-phase roles should be reduced before production. | A policy generated from the function's activity. | | Long-term IAM user keys inside a workload | AWS says to use roles so no long-lived credentials are distributed to workloads on AWS. | The execution, instance or task role. | | One shared node role for every pod | AWS says that if metadata access is not restricted, pod containers can reach the node role's credentials. | Pod Identity or IRSA, plus restricted metadata access. | | Pods with hostNetwork: true | AWS says these pods always have metadata service access. | Avoid the setting unless the pod truly needs it. | | Manually calling sts:AssumeRole for a Lambda execution role | AWS says to avoid it and that self-assumption requires the role in its own trust policy. | Let Lambda assume the role at invocation. | ## Measure actions granted against actions called per role For each workload role, record the actions in its policies and the actions in CloudTrail over the window, then report the share granted but unused. Track roles whose policies contain a wildcard action as a separate count, because generated policies remove those first. Count pods or functions that still run under a shared role. A falling count means each workload is moving to its own identity. [The principle of least privilege](/principle-of-least-privilege/) explains why an unused permission is exposure, and [privilege creep](/privilege-creep/) shows how a role grows between releases. Related guides: [Azure workload identities](/guides/azure-workload-identities/), [Google Cloud workload identities](/guides/google-cloud-workload-identities/) and [AWS third-party identities](/guides/aws-third-party-identities/). ## Related pages - [Least privilege for third-party identities in AWS](/guides/aws-third-party-identities/) - [Least privilege for human users in Azure](/guides/azure-human-identities/) - [Least privilege for service identities in Azure](/guides/azure-service-identities/) - [All of them](/guides/) ## Sources - AWS IAM User Guide: Security best practices in IAM: https://docs.aws.amazon.com/IAM/latest/UserGuide/best-practices.html (read 16 Nov 2026) - AWS Lambda Developer Guide: Lambda execution role: https://docs.aws.amazon.com/lambda/latest/dg/lambda-intro-execution-role.html (read 16 Nov 2026) - Amazon EKS User Guide: IAM roles for service accounts: https://docs.aws.amazon.com/eks/latest/userguide/iam-roles-for-service-accounts.html (read 16 Nov 2026) - Amazon EKS User Guide: EKS Pod Identity: https://docs.aws.amazon.com/eks/latest/userguide/pod-identities.html (read 16 Nov 2026) - AWS IAM User Guide: IAM Access Analyzer policy generation: https://docs.aws.amazon.com/IAM/latest/UserGuide/access-analyzer-policy-generation.html (read 16 Nov 2026) - AWS IAM User Guide: IAM Access Analyzer findings: https://docs.aws.amazon.com/IAM/latest/UserGuide/access-analyzer-findings.html (read 16 Nov 2026) - AWS IAM User Guide: Refine permissions using last accessed information: https://docs.aws.amazon.com/IAM/latest/UserGuide/access_policies_last-accessed.html (read 16 Nov 2026) - AWS IAM User Guide: AWS managed policies for job functions: https://docs.aws.amazon.com/IAM/latest/UserGuide/access_policies_job-functions.html (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/guides/aws-workload-identities/ --- # Least privilege for third-party identities in AWS means narrow cross-account roles with an external ID Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. Third-party identities in AWS are the cross-account roles that vendors, partners and your own other accounts assume. AWS's IAM documentation, read 16 Nov 2026, describes the trust policy, the external ID and confused deputy conditions. This guide lists the tools, steps, policies to avoid and measures. ## Third parties reach an AWS account by assuming a role you create AWS's documentation on third-party access says you can delegate access by creating a role that the third party assumes, so no AWS security credentials are shared. The third party supplies its AWS account ID, which you name as the principal in the trust policy. The role has two parts. The trust policy says who may assume it. The permissions policy says what the assumed role can do. AWS states that a third party can access any resource you specify in the policy, and that their use of your resources is billed to you, so the policy is where you limit them. A second case is your own accounts. AWS documents granting IAM users permission to switch to roles in other accounts you own, for example a production account where administrators assume a role only when they must terminate an instance. The same cross-account structure applies, with you on both sides. A third case is AWS services acting across accounts, for example CloudTrail writing to a bucket in another account. AWS calls the risk in both the vendor case and the service case the confused deputy problem. ## The external ID and source conditions are AWS's documented controls for third-party roles AWS says the external ID is an identifier known only to you and the third party, that you set it in the trust policy through the sts:ExternalId condition, and that its primary function is to prevent the confused deputy problem. AWS tells you not to use something guessable, such as the third party's name or phone number. It also states that AWS does not treat an external ID as a secret and anyone who can view the role can see it. The confused deputy documentation describes how, without an external ID, another customer of the same vendor could supply your role ARN, which is not a secret, and have the vendor act in your account. For service-to-service access, AWS recommends the aws:SourceArn, aws:SourceAccount, aws:SourceOrgID and aws:SourceOrgPaths condition keys in resource policies. For finding what third parties can actually reach, IAM Access Analyzer external access analyzers identify resources and roles shared with an external entity, and AWS says external access findings are free. The unused access analyzer then shows whether a vendor role is idle or holds permissions its caller never uses; AWS states unused access findings are charged. Neither tool replaces knowing who the vendor is, which [the method page](/methodology/) treats as an inventory problem before a tooling problem. ## Steps to bring third-party and cross-account roles to least privilege - Run an external access analyzer for the organization and list every role that principals outside your zone of trust can assume. - For each role, record the vendor, the business owner, the purpose and the date access was last needed. - Confirm every vendor role has an sts:ExternalId condition, and replace guessable values with unique ones per customer relationship. - Replace AdministratorAccess or other broad managed policies with a policy that lists the vendor's required actions on named resources. Ask the vendor for its required permission list; AWS says the third party must tell you the permissions it needs. - Generate a policy from CloudTrail activity for the role (up to 90 days) and compare it with what the vendor documented. - Add conditions where the vendor can meet them. AWS's best practices page gives requiring TLS as one example of a policy condition. - Delete roles that the unused access analyzer reports as unused, after notifying the owner. - Set a review date per vendor and revoke the role when the contract ends. ## Policies and trust settings to avoid on third-party roles | Policy or setting | Documented reason (AWS, read 16 Nov 2026) | Use instead | | --- | --- | --- | | A vendor role without an external ID | AWS says the external ID prevents the confused deputy problem when a vendor serves many customers. | A unique external ID in the trust policy condition. | | A guessable external ID | AWS says not to use something that can be guessed, such as a company name or phone number. | An identifier known only to you and the vendor. | | AdministratorAccess on a vendor role | AWS describes it as all actions on all resources, and says third parties can reach any resource the policy names. | A policy of named actions on named resources. | | Vendor roles nobody uses | AWS says the unused access analyzer reports roles with no access activity in the usage window. | Delete the role after notifying the owner. | | Service principal access without source conditions | AWS says a bucket trusting a service principal with no conditions can receive data from an unauthorized actor's account. | aws:SourceArn or aws:SourceAccount conditions. | ## Measure vendor roles, their age and the share of their permissions used Count the roles assumable from outside your organization, the share with an external ID, and the share that have an assigned owner. Then compare, per vendor role, the actions granted with the actions seen in CloudTrail. A large gap between actions granted and actions seen is the clearest signal. Roles with no activity at all are the next signal, because they are access nobody is using and nobody is watching. [Privilege creep](/privilege-creep/) describes how such roles accumulate, and [the glossary](/glossary/) defines the terms used above. Related guides: [Azure third-party identities](/guides/azure-third-party-identities/) and [Google Cloud third-party identities](/guides/google-cloud-third-party-identities/). ## Related pages - [Least privilege for human users in Azure](/guides/azure-human-identities/) - [Least privilege for service identities in Azure](/guides/azure-service-identities/) - [Least privilege for workload identities in Azure](/guides/azure-workload-identities/) - [All of them](/guides/) ## Sources - AWS IAM User Guide: Access to AWS accounts owned by third parties: https://docs.aws.amazon.com/IAM/latest/UserGuide/id_roles_common-scenarios_third-party.html (read 16 Nov 2026) - AWS IAM User Guide: The confused deputy problem: https://docs.aws.amazon.com/IAM/latest/UserGuide/confused-deputy.html (read 16 Nov 2026) - AWS IAM User Guide: Access for an IAM user in another AWS account that you own: https://docs.aws.amazon.com/IAM/latest/UserGuide/id_roles_common-scenarios_aws-accounts.html (read 16 Nov 2026) - AWS IAM User Guide: Using IAM Access Analyzer: https://docs.aws.amazon.com/IAM/latest/UserGuide/what-is-access-analyzer.html (read 16 Nov 2026) - AWS IAM User Guide: IAM Access Analyzer findings: https://docs.aws.amazon.com/IAM/latest/UserGuide/access-analyzer-findings.html (read 16 Nov 2026) - AWS IAM User Guide: IAM Access Analyzer policy generation: https://docs.aws.amazon.com/IAM/latest/UserGuide/access-analyzer-policy-generation.html (read 16 Nov 2026) - AWS IAM User Guide: AWS managed policies for job functions: https://docs.aws.amazon.com/IAM/latest/UserGuide/access_policies_job-functions.html (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/guides/aws-third-party-identities/ --- # Least privilege for human users in Azure means eligible role assignments to groups at narrow scopes Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. Human users in Azure get permissions from Entra users and groups through Azure RBAC role assignments at a scope. Microsoft's documentation, read 16 Nov 2026, names PIM, access reviews and Defender for Cloud recommendations as the provider tools. This guide gives the steps, the roles to avoid and the measures. ## People get Azure permissions from role assignments made to users and groups at a scope An Azure RBAC role assignment ties a security principal, a role definition and a scope. Microsoft's RBAC best practices page says to grant users the least privilege to get their work done, to avoid broader roles at broader scopes even when that looks convenient, and to assign roles to groups rather than directly to users. The page gives two reasons for groups. Assignments are easier to manage, and there is a limit on role assignments per subscription, which groups help stay under. Microsoft Entra roles are a separate system from Azure RBAC roles. Entra roles such as Global Administrator control the directory; Azure roles control subscriptions and resources. Microsoft's access review documentation lists Global Administrators and Azure resource roles together as role assignments worth recertifying. Privilege accumulates through [privilege creep](/privilege-creep/): a person joins a project, gets Contributor on a subscription, and keeps it after moving teams. ## Microsoft provides PIM, access reviews and Defender for Cloud recommendations Microsoft Entra Privileged Identity Management (PIM) provides time-based and approval-based role activation for Entra roles and Azure resources. Microsoft lists time-bound assignments with start and end dates, approval to activate, multifactor authentication on activation, justification, notifications and audit history. Microsoft says an eligible role assignment gives the same access as a permanent one when activated; the difference is that access is not standing. Microsoft states that PIM requires licenses. Access reviews in Microsoft Entra let reviewers approve or deny continued access for group memberships, enterprise applications and role assignments, recurring weekly, monthly, quarterly or annually. Microsoft notes the feature requires Microsoft Entra ID Governance or Microsoft Entra Suite subscriptions, with some capabilities available under Microsoft Entra ID P2. Defender for Cloud includes cloud infrastructure entitlement management capabilities in its Defender Cloud Security Posture Management plan. Microsoft says it surfaces recommendations such as removing inactive, guest or blocked accounts with access, limiting administrative privileges to a defined set of users, and right-sizing permissions for overprovisioned identities based on actual usage. Microsoft's RBAC page adds a monitored count: have a maximum of three subscription owners, a recommendation that Defender for Cloud monitors. It also says to check privileged administrator role assignments and, when you assign a role that can create role assignments, to consider a condition that constrains it. ## Steps to bring human access in Azure to least privilege - List role assignments for every subscription and management group, filtered to privileged administrator roles, and record principal, scope and assignment date. - Move direct user assignments to Microsoft Entra security groups, one group per role and scope. - Convert standing assignments of Owner, User Access Administrator, Role Based Access Control Administrator and Contributor to eligible assignments in PIM, with activation requiring MFA and justification. - Reduce subscription owners to three or fewer and narrow remaining privileged assignments to resource group or resource scope. - Create recurring access reviews for the groups behind privileged roles, with the group owner or the manager as reviewer. - Work through the Defender for Cloud recommendations on inactive accounts and overprovisioned permissions, starting with identities that can reach sensitive resources. - Replace wildcard (*) custom roles with explicit Actions and DataActions, as Microsoft advises. - Repeat the review on the cadence the access review sets. [The first 90 days](/first-90-days/) gives an order for these steps. ## Roles and assignments to avoid for human users | Role or assignment | Documented reason (Microsoft, read 16 Nov 2026) | Use instead | | --- | --- | --- | | Owner | Microsoft describes it as full access to manage all resources, including the ability to assign roles in Azure RBAC. | A job-function role at resource group scope; Owner eligible through PIM. | | User Access Administrator | Microsoft describes it as letting you manage user access to Azure resources. | Role Based Access Control Administrator with a condition, or an eligible assignment. | | Contributor at subscription scope | Microsoft advises avoiding broader roles at broader scopes and prefers job-function roles to privileged administrator roles. | A narrower built-in role at resource group scope. | | More than three subscription owners | Microsoft says to have a maximum of three, to reduce the chance of breach by a compromised owner. | Eligible owner assignments through PIM. | | Custom roles with wildcard actions | Microsoft says future Actions or DataActions added under a wildcard may grant unwanted access. | Explicitly listed Actions and DataActions. | | Roles assigned directly to individual users | Microsoft says to assign to groups to keep assignments manageable. | Security groups. | ## Measure standing privileged assignments and the share of permissions used Count standing (permanent) assignments of privileged administrator roles against eligible ones. A move from standing to eligible is visible in PIM without any usage data. Count subscriptions with more than three owners, and count users with administrative roles who have not signed in during the review window. For overprovisioning, use the Defender for Cloud right-sizing recommendations, which compare granted permissions with actual usage, and record their number over time. The [Access Debt Calculator](/) accepts granted and used counts, and the [policy template](/least-privilege-policy-template/) offers wording for the policy that sets these thresholds. Related guides: [AWS human identities](/guides/aws-human-identities/) and [Google Cloud human identities](/guides/google-cloud-human-identities/). ## Related pages - [Least privilege for service identities in Azure](/guides/azure-service-identities/) - [Least privilege for workload identities in Azure](/guides/azure-workload-identities/) - [Least privilege for third-party identities in Azure](/guides/azure-third-party-identities/) - [All of them](/guides/) ## Sources - Microsoft Learn: Best practices for Azure RBAC: https://learn.microsoft.com/en-us/azure/role-based-access-control/best-practices (read 16 Nov 2026) - Microsoft Learn: Azure built-in roles for Privileged: https://learn.microsoft.com/en-us/azure/role-based-access-control/built-in-roles/privileged (read 16 Nov 2026) - Microsoft Learn: What is Microsoft Entra Privileged Identity Management: https://learn.microsoft.com/en-us/entra/id-governance/privileged-identity-management/pim-configure (read 16 Nov 2026) - Microsoft Learn: What are access reviews: https://learn.microsoft.com/en-us/entra/id-governance/access-reviews-overview (read 16 Nov 2026) - Microsoft Learn: Permissions management (CIEM) in Defender for Cloud: https://learn.microsoft.com/en-us/azure/defender-for-cloud/permissions-management (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/guides/azure-human-identities/ --- # Least privilege for service identities in Azure means scoping service principals and retiring their secrets Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. Service identities in Azure are service principals created from app registrations, used by pipelines, scripts and applications. Microsoft's documentation, read 16 Nov 2026, covers credentials, Conditional Access for workload identities and Defender for Cloud recommendations. This guide gives the steps, the roles to avoid and what to measure. ## A service principal is the tenant-local identity of an app registration and carries its own role assignments Microsoft's documentation defines an application object as the global representation of an app and a service principal as its local representation in a specific tenant. The service principal defines what the app can do in that tenant and what resources it can access. Microsoft names three types: application, managed identity and legacy. After app registration you can add secrets or certificates to make the app work. Microsoft's workload identity overview says a workload may deal with multiple credentials to access different resources, those credentials must be stored securely, and it is hard to track when a workload identity is created or when it should be revoked. Permissions come from Azure RBAC role assignments to the service principal, and for Microsoft Graph and other APIs from consented application permissions. This guide covers the first; the second follows the same logic of requesting only what the app calls. Managed identities are a special type of service principal and are covered in the [workload identities guide](/guides/azure-workload-identities/). The remaining service principals are the ones that hold secrets, and they carry most of the exposure. ## Microsoft's tools for service principals are Conditional Access, ID Protection and Defender for Cloud Microsoft Entra Workload ID supports Conditional Access policies for service principals owned by your organization, and continuous access evaluation for workload identities. Microsoft says Microsoft Entra ID Protection can detect risks such as leaked credentials for workload identities. Defender for Cloud assesses service principals, groups, managed identities and service accounts for excessive, unused or misconfigured permissions, according to Microsoft. It lists right-sizing recommendations for overprovisioned identities based on actual usage and attack path analysis that can show lateral movement from an overprivileged service principal. Microsoft's RBAC best practices apply to service principals as they do to users: narrow scope, no wildcard custom roles, and a condition when the role can create role assignments. Microsoft's [privileged roles reference](https://learn.microsoft.com/en-us/azure/role-based-access-control/built-in-roles/privileged) lists Owner, User Access Administrator and Role Based Access Control Administrator among the roles to treat as privileged administrator roles. Access reviews in Entra include role assignments as a review type, and the [routes page](/routes/) shows where provider tools stop and other routes begin. ## Steps to bring service principals to least privilege - List all service principals with Azure role assignments, with owner, creation date, credential type and credential expiry. - Identify service principals that run on Azure compute and replace them with managed identities, which remove stored credentials. - For pipelines outside Azure, replace secrets with workload identity federation where the platform supports it (see the workload guide). - For the service principals that must keep a credential, prefer certificates to secrets, set the shortest expiry the owner can support and record the owner. - Replace Owner, Contributor and User Access Administrator assignments with the narrowest built-in role at resource group or resource scope. - Where a service principal can create role assignments, add a condition so it can assign only the roles you list. - Apply Conditional Access for workload identities to service principals that sign in from known locations. - Delete service principals with no sign-in activity after a notice period, and repeat quarterly. ## Roles and credentials to avoid on service principals | Role or credential | Documented reason (Microsoft, read 16 Nov 2026) | Use instead | | --- | --- | --- | | Owner on a subscription | Microsoft describes it as full access including the ability to assign roles in Azure RBAC. | A narrow built-in role at resource group scope. | | Contributor on a subscription | Microsoft describes it as full access to manage all resources; Microsoft advises avoiding privileged administrator roles where a job-function role can be used. | A resource-specific role such as a data or deployment role. | | User Access Administrator | Microsoft describes it as letting you manage user access to Azure resources. | Role Based Access Control Administrator with a condition on the roles it may assign. | | Client secrets with no owner or expiry plan | Microsoft says credentials need secure storage and that it is hard to track when to revoke a workload identity. | Managed identity or federated credential; otherwise a certificate with a recorded owner. | | Wildcard custom roles | Microsoft says wildcards can grant future Actions or DataActions that you did not intend. | Explicit Actions and DataActions. | ## Measure service principals by credential age, scope and used permissions Count service principals that hold a secret, the share that have an owner, and the share whose credentials expire within a set period. Count service principals with Owner, Contributor or User Access Administrator, and the share assigned at subscription or management group scope. For the used side, track the Defender for Cloud right-sizing recommendations that name service principals, and sign-in activity from the Entra sign-in logs for service principals. A principal with a broad role and no sign-in in the review window is the first deletion candidate. [The principle of least privilege](/principle-of-least-privilege/) explains why an idle principal still counts. Related guides: [AWS service identities](/guides/aws-service-identities/) and [Google Cloud service identities](/guides/google-cloud-service-identities/). ## Related pages - [Least privilege for workload identities in Azure](/guides/azure-workload-identities/) - [Least privilege for third-party identities in Azure](/guides/azure-third-party-identities/) - [Least privilege for human users in Google Cloud](/guides/google-cloud-human-identities/) - [All of them](/guides/) ## Sources - Microsoft Learn: Application and service principal objects: https://learn.microsoft.com/en-us/entra/identity-platform/app-objects-and-service-principals (read 16 Nov 2026) - Microsoft Learn: Workload identities: https://learn.microsoft.com/en-us/entra/workload-id/workload-identities-overview (read 16 Nov 2026) - Microsoft Learn: Best practices for Azure RBAC: https://learn.microsoft.com/en-us/azure/role-based-access-control/best-practices (read 16 Nov 2026) - Microsoft Learn: Azure built-in roles for Privileged: https://learn.microsoft.com/en-us/azure/role-based-access-control/built-in-roles/privileged (read 16 Nov 2026) - Microsoft Learn: Permissions management (CIEM) in Defender for Cloud: https://learn.microsoft.com/en-us/azure/defender-for-cloud/permissions-management (read 16 Nov 2026) - Microsoft Learn: What are access reviews: https://learn.microsoft.com/en-us/entra/id-governance/access-reviews-overview (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/guides/azure-service-identities/ --- # Least privilege for workload identities in Azure means managed identities and federated credentials with narrow role assignments Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. Workload identities in Azure are managed identities and federated service principals used by software on Azure compute, Kubernetes and CI/CD. Microsoft's documentation, read 16 Nov 2026, explains system-assigned and user-assigned identities and workload identity federation. This guide covers cutting their roles and measuring use. ## Managed identities and federated credentials replace stored secrets for workloads Microsoft's documentation says managed identities remove the need for developers to manage credentials, because the credentials are not accessible to them. A managed identity can be assigned to Azure compute such as virtual machines, virtual machine scale sets, Service Fabric clusters, AKS and supported app hosting platforms. Microsoft states managed identities can be used at no extra cost. There are two kinds. A system-assigned identity is created with the resource and deleted with it, and only that resource can request tokens with it. A user-assigned identity is a standalone resource that can be assigned to several resources, and Microsoft calls it the recommended type for Microsoft services. Workload identity federation lets a user-assigned managed identity or app registration trust tokens from an external identity provider. Microsoft lists GitHub Actions, any Kubernetes cluster (AKS, EKS, GKE or on-premises), Google Cloud, AWS, SPIFFE and Azure Pipelines service connections as supported scenarios. The federated identity credential states which token from the provider is trusted. Both mechanisms remove the secret. Neither limits what the identity can do. That is set by its Azure role assignments. ## Microsoft's best practice page shows where convenience works against least privilege Microsoft's managed identity recommendations say user-assigned identities suit more scenarios and reduce administration. In its example, four virtual machines sharing one user-assigned identity need two role assignments, against eight with system-assigned identities. The cost is that permissions are shared. Microsoft states that when an identity needs more role assignments they are granted to all the resources associated with it. If each resource needs a unique set of permissions, Microsoft says to use a system-assigned identity, and it also recommends system-assigned identities when you need the permissions removed along with the resource or need logs that show which resource acted. Microsoft's best practices also allow a resource to have both a system-assigned and one or more user-assigned identities, so shared access and granular access can be separated. Defender for Cloud covers managed identities in its permission analysis, with recommendations for overprovisioned identities based on actual usage, as Microsoft documents. The Azure RBAC rules from the [human identities guide](/guides/azure-human-identities/) apply: narrow scope, no wildcards. ## Steps to bring workload identities to least privilege - List managed identities and the resources that use them. For each user-assigned identity, record every attached resource. - Split any shared user-assigned identity whose attached resources need different access, or move those resources to system-assigned identities. - Replace app registrations that hold secrets with managed identities for workloads running on Azure. - For GitHub Actions, Kubernetes and other external workloads, create federated identity credentials in place of secrets and restrict each credential to the issuer and subject it should trust. - Replace Owner and Contributor assignments with data-plane or job-specific roles at resource scope. - Remove role assignments for deleted resources and for identities that no resource uses. - Review Defender for Cloud right-sizing recommendations for managed identities and apply them in a non-production subscription first. - Set a review cadence. See [what least privilege access covers](/least-privilege-access/) for the ordering. ## Roles and patterns to avoid on workload identities | Role or pattern | Documented reason (Microsoft, read 16 Nov 2026) | Use instead | | --- | --- | --- | | Owner or Contributor on a managed identity | Microsoft describes Owner as including role assignment and Contributor as full resource management. | A role scoped to the single downstream service. | | One user-assigned identity shared by unlike workloads | Microsoft says added role assignments are granted to every resource using the identity. | Identities per permission set. | | Client secrets for workloads that can federate | Microsoft says federation removes secrets for GitHub, Kubernetes and other external providers. | Federated identity credentials. | | Broad federated trust (any subject) | Microsoft says the federated credential defines which token from the provider is trusted. | A credential limited to the exact issuer and subject. | | Wildcard custom roles | Microsoft says wildcards can add future permissions unintentionally. | Explicit Actions and DataActions. | ## Measure assignments per identity and use per assignment Record the count of role assignments per managed identity and the count of resources attached to each user-assigned identity. A rising attached-resource count with a flat role list means the identity's permissions are shared more widely than when it was designed. Compare granted roles with use through the Defender for Cloud right-sizing recommendations and the Azure Activity Log. Count external credentials by type: secret, certificate, federated. The trend toward federated is the simplest sign of progress. [Privilege creep](/privilege-creep/) describes why the role list rarely shrinks without this measurement. Related guides: [AWS workload identities](/guides/aws-workload-identities/) and [Google Cloud workload identities](/guides/google-cloud-workload-identities/). ## Related pages - [Least privilege for third-party identities in Azure](/guides/azure-third-party-identities/) - [Least privilege for human users in Google Cloud](/guides/google-cloud-human-identities/) - [Least privilege for service identities in Google Cloud](/guides/google-cloud-service-identities/) - [All of them](/guides/) ## Sources - Microsoft Learn: What are managed identities for Azure resources: https://learn.microsoft.com/en-us/entra/identity/managed-identities-azure-resources/overview (read 16 Nov 2026) - Microsoft Learn: Managed identity best practice recommendations: https://learn.microsoft.com/en-us/entra/identity/managed-identities-azure-resources/managed-identity-best-practice-recommendations (read 16 Nov 2026) - Microsoft Learn: Workload identity federation: https://learn.microsoft.com/en-us/entra/workload-id/workload-identity-federation (read 16 Nov 2026) - Microsoft Learn: Workload identities: https://learn.microsoft.com/en-us/entra/workload-id/workload-identities-overview (read 16 Nov 2026) - Microsoft Learn: Permissions management (CIEM) in Defender for Cloud: https://learn.microsoft.com/en-us/azure/defender-for-cloud/permissions-management (read 16 Nov 2026) - Microsoft Learn: Best practices for Azure RBAC: https://learn.microsoft.com/en-us/azure/role-based-access-control/best-practices (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/guides/azure-workload-identities/ --- # Least privilege for third-party identities in Azure means narrow Lighthouse authorizations and reviewed multitenant app consent Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. Third-party identities in Azure are the partner users, delegated Lighthouse authorizations and multitenant application service principals that act in your tenant. Microsoft's documentation, read 16 Nov 2026, lists which roles Lighthouse supports and how consent creates service principals. This guide covers the steps and measures. ## Vendors reach Azure through Lighthouse delegation, multitenant app consent or guest accounts Azure Lighthouse lets a service provider manage resources in a customer's tenant from the provider's own tenant. Microsoft says customers keep control over the scopes they delegate and the permissions allowed, can audit provider actions and can remove access at any time. Delegated users can be seen in the customer's activity log. Each Lighthouse authorization pairs a principalId in the managing tenant (user, group or service principal) with a built-in role. Microsoft states Lighthouse does not support custom roles, does not support the Owner role, supports User Access Administrator only to assign roles to a managed identity in the customer tenant, and does not support roles with DataActions or with several Microsoft.Authorization write and delete actions. A multitenant application is registered once and, when a user or admin in another tenant consents, a service principal is created in that tenant. Microsoft says app-only permissions always require tenant administrator consent, and that an administrator can revoke the app's access in Enterprise applications. Microsoft Entra B2B collaboration represents partners as guest user objects in your directory, using their own credentials. ## Microsoft documents scope, group-based delegation and eligible authorizations as the controls Microsoft's recommended practices for Lighthouse say to enable the minimum access needed, assign permissions to security groups using the principle of least privilege, review membership regularly and require multifactor authentication in both tenants. Microsoft states that Conditional Access policies on a customer tenant do not apply to users who reach the customer's resources through Lighthouse; only policies set on the managing tenant apply. Microsoft also says that when you onboard through a public managed service offer, every group, user or service principal in the plan has the same permissions for every customer who buys it. Separate access needs a private plan per customer or onboarding by Azure Resource Manager template. Eligible authorizations let users in the managing tenant temporarily activate a higher role instead of holding permanent privileged assignments; Microsoft notes licensing requirements. Microsoft also advises including an authorization with the Managed Services Registration Assignment Delete Role, so the provider can remove the delegation. Otherwise only a user in the customer's tenant can remove it. For guests, access reviews can ask group owners to confirm that guests still have a business need, as Microsoft describes, and the [methodology page](/methodology/) explains how this site treats such provider tools. ## Steps to bring third-party access in Azure to least privilege - List Lighthouse delegations (the Service providers page in the Azure portal), enterprise applications from other tenants, and guest users with Azure role assignments. - For each delegation, record the provider, scope, roles and the business owner. Remove delegations with no owner. - Replace broad built-in roles in an authorization with the narrowest built-in role that does the job, at resource group scope where possible. - Ask providers to use Entra groups and eligible authorizations, and to require MFA for every user in their managing tenant. - Review multitenant application service principals in Enterprise applications: check consented permissions, remove those the app no longer uses and revoke apps that are not in use. - Turn off end-user consent where a tenant administrator should approve third-party apps, using the Consent and permissions setting Microsoft documents. - Create recurring access reviews for guest users and for groups with guests. - End each delegation when the contract ends. ## Roles and settings to avoid for third parties | Role or setting | Documented reason (Microsoft, read 16 Nov 2026) | Use instead | | --- | --- | --- | | Owner in a Lighthouse authorization | Microsoft states Lighthouse does not support the Owner role. | A narrower built-in role. | | Contributor for a provider that needs read access | Microsoft says to enable only the minimum access necessary to manage the resources. | Reader or a service-specific role. | | A public plan with one group for all customers | Microsoft says every group in a public plan has the same permissions for every customer who buys it. | A private plan per customer or ARM template onboarding. | | Relying on customer Conditional Access for provider users | Microsoft says those policies do not apply to users coming through Lighthouse. | MFA enforced in the managing tenant too. | | Applications with unreviewed app-only permissions | Microsoft says app-only permissions require tenant administrator consent and can be revoked in Enterprise applications. | Reviewed consent and revocation of unused apps. | ## Measure delegations, consented permissions and use Count active Lighthouse delegations, the roles each carries, and the share with an assigned business owner. Count multitenant applications with consented permissions and the number of permissions per application. Compare granted permissions with the permissions the application's sign-ins and activity show. A delegation or app with no recorded activity in the review window is the first candidate for removal. See [privilege creep](/privilege-creep/) for how these accumulate, and [the glossary](/glossary/) for definitions. Related guides: [AWS third-party identities](/guides/aws-third-party-identities/) and [Google Cloud third-party identities](/guides/google-cloud-third-party-identities/). ## Related pages - [Least privilege for human users in Google Cloud](/guides/google-cloud-human-identities/) - [Least privilege for service identities in Google Cloud](/guides/google-cloud-service-identities/) - [Least privilege for workload identities in Google Cloud](/guides/google-cloud-workload-identities/) - [All of them](/guides/) ## Sources - Microsoft Learn: What is Azure Lighthouse: https://learn.microsoft.com/en-us/azure/lighthouse/overview (read 16 Nov 2026) - Microsoft Learn: Tenants, users, and roles in Azure Lighthouse scenarios: https://learn.microsoft.com/en-us/azure/lighthouse/concepts/tenants-users-roles (read 16 Nov 2026) - Microsoft Learn: Recommended security practices for Azure Lighthouse: https://learn.microsoft.com/en-us/azure/lighthouse/concepts/recommended-security-practices (read 16 Nov 2026) - Microsoft Learn: Convert single-tenant app to multitenant: https://learn.microsoft.com/en-us/entra/identity-platform/howto-convert-app-to-be-multi-tenant (read 16 Nov 2026) - Microsoft Learn: What is Microsoft Entra B2B collaboration: https://learn.microsoft.com/en-us/entra/external-id/what-is-b2b (read 16 Nov 2026) - Microsoft Learn: What are access reviews: https://learn.microsoft.com/en-us/entra/id-governance/access-reviews-overview (read 16 Nov 2026) - Microsoft Learn: Azure built-in roles for Privileged: https://learn.microsoft.com/en-us/azure/role-based-access-control/built-in-roles/privileged (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/guides/azure-third-party-identities/ --- # Least privilege for human users in Google Cloud means predefined roles for groups at the smallest scope Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. Human users in Google Cloud get permissions when IAM roles are granted to users and groups in allow policies. Google's documentation, read 16 Nov 2026, names the IAM recommender, Policy Troubleshooter, Policy Analyzer and Policy Simulator as the provider tools. This guide gives the steps, the roles to avoid and the measures. ## People get Google Cloud permissions from roles granted in allow policies An IAM allow policy binds principals to roles on a resource, and Google's documentation says child resources inherit the allow policies of their parents. Its example: if a project policy lets a user administer Compute Engine VMs, the user can administer any VM in the project, whatever is set on an individual VM. Google's IAM security guidance therefore says to grant roles at the smallest scope needed and to grant roles to groups instead of individual users when possible. Google describes three kinds of role. Basic roles (Owner, Editor, Viewer, with newer Admin, Writer and Reader equivalents) are broad. Predefined roles are service-specific. Custom roles are ones you define. Google says basic roles include thousands of permissions across all services and that in production you should not grant them unless there is no alternative. Google also says project creators are automatically granted the Owner role on the new project. That is the first place [privilege creep](/privilege-creep/) starts in Google Cloud. ## The IAM recommender, Policy Troubleshooter and Policy Simulator are Google's tools for this identity type The IAM recommender produces role recommendations that suggest removing or replacing roles that give excess permissions. Google says it compares a principal's permissions with the permissions it used in the last 90 days, or since the grant if the role was granted fewer than 90 days ago. It also uses a machine learning model to keep permissions the principal is likely to need, which Google marks with a distinct symbol in the comparison view. Policy Intelligence is the name Google gives to a set of tools. Policy Analyzer for allow policies finds which principals have access to which resources. Policy Troubleshooter takes a principal, a resource and a permission and examines the allow, deny and Principal Access Boundary policies to say whether access is allowed. Policy Simulator shows how a policy or role change would affect access before you make it. Google's IAM security guidance suggests using the recommender and Policy Simulator when replacing a basic role, and conditional role bindings so access expires automatically, with temporary elevated access for administrators. Google's documentation states that recommendations are not generated for every role granted to a principal, so absence of a recommendation does not show a role is right-sized. ## Steps to bring human access in Google Cloud to least privilege - Use Policy Analyzer to list who holds roles/owner, roles/editor and other basic roles across the organization, and at which level. - Move individual users into Google groups and grant roles to the groups. - For every basic role, review the IAM recommender suggestion and the role's predefined replacement. Test the swap in Policy Simulator before applying. - Move grants from the organization or folder to the project or resource that needs them. - Convert standing admin grants to time-limited conditional bindings or temporary access. - Limit who can modify allow policies: Google notes that roles such as Project IAM Admin and Folder IAM Admin let holders change policies without direct resource access. - Use Policy Troubleshooter when a removal causes a denied request, to see which policy applies. - Reapply the recommender review on a fixed cycle. [The first 90 days](/first-90-days/) suggests an order. ## Roles to avoid for human users | Role | Documented reason (Google, read 16 Nov 2026) | Use instead | | --- | --- | --- | | roles/owner | Google says it allows access to and modification of almost all resources, including allow policies, and to grant it only when nearly universal access is required. | A predefined admin role for one service, at project scope. | | roles/editor | Google says it lets you create and delete resources for most services; it is a legacy basic role that cannot take IAM Conditions. | A predefined role for the services the person uses. | | roles/viewer | Google says basic roles include thousands of permissions across all services. | A service-specific viewer role. | | Service Account User on a project | Google says holders can access every resource the service account can access. | The role on one service account only. | | Project IAM Admin or Folder IAM Admin for routine tasks | Google says these roles allow modifying allow policies. | Narrow admin roles; grant these through temporary access. | | Any role at organization scope when one project needs it | Google says allow policies inherit down the hierarchy. | Grant at the project or resource. | ## Measure permissions granted against permissions used in the last 90 days The IAM recommender's own window gives the unit. For each human principal and role, record the number of permissions in the role and the number the principal used in the last 90 days, and keep the unused share per role. Record the count of principals with basic roles, split by Owner and Editor, and the count of grants made directly to users rather than groups. These counts are available without any new tool: Policy Analyzer gives the first two and Cloud Asset data the third. The [Access Debt Calculator](/) turns granted and used counts into a single figure. [The principle of least privilege](/principle-of-least-privilege/) explains why the unit is the permission. Related guides: [AWS human identities](/guides/aws-human-identities/) and [Azure human identities](/guides/azure-human-identities/). ## Related pages - [Least privilege for service identities in Google Cloud](/guides/google-cloud-service-identities/) - [Least privilege for workload identities in Google Cloud](/guides/google-cloud-workload-identities/) - [Least privilege for third-party identities in Google Cloud](/guides/google-cloud-third-party-identities/) - [All of them](/guides/) ## Sources - Google Cloud IAM: Best practices for using IAM securely: https://cloud.google.com/iam/docs/using-iam-securely (read 16 Nov 2026) - Google Cloud IAM: Roles and permissions: https://cloud.google.com/iam/docs/roles-overview (read 16 Nov 2026) - Google Cloud IAM: Role recommendations overview: https://cloud.google.com/iam/docs/recommender-overview (read 16 Nov 2026) - Google Cloud IAM: Manage role recommendations: https://cloud.google.com/iam/docs/recommender-managing (read 16 Nov 2026) - Google Cloud: Policy Intelligence overview: https://cloud.google.com/policy-intelligence/docs/overview (read 16 Nov 2026) - Google Cloud IAM: Troubleshoot permissions: https://cloud.google.com/iam/docs/troubleshooting-access (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/guides/google-cloud-human-identities/ --- # Least privilege for service identities in Google Cloud means dedicated service accounts, no keys and no Editor role Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. Service identities in Google Cloud are service accounts and Google-managed service agents. Google's documentation, read 16 Nov 2026, says default service accounts can hold the Editor role and that service account keys are a security risk. This guide covers cutting them with the IAM recommender and usage tools. ## Service accounts, their keys and service agents are three different exposures Google's service account guidance says some services, including Compute Engine and App Engine, create a default service account that has the Editor role (roles/editor) on the project by default. A resource created without a named service account can use it automatically. Google calls it very risky to share a powerful service account across multiple applications, and notes shared accounts make it hard to trace activity to the right application. Service account keys are user-managed credentials. Google lists the main threats as credential leakage, privilege escalation, exposure of confidential metadata, non-repudiation and malicious credential configurations, and says the best mitigation is to avoid user-managed keys. Service agents are Google-managed service accounts that let a service access your resources. Google says it creates them after you activate and use an API, and they appear in allow policies and audit logs. You do not create them, but you can see and review their roles. Google's IAM guidance adds that anyone granted the Service Account User role on a service account can access all the resources that service account can access. ## Recommender, usage views and lateral movement insights are the provider tools Google's Policy Intelligence tools for service accounts include viewing recent usage for service accounts and keys, finding unused service accounts and finding service accounts with lateral movement permissions. Google recommends identifying and disabling unused service accounts, and disabling before deleting. The IAM recommender covers service accounts through role recommendations built from the last 90 days of permission use. Google notes that role recommendations for service agents can include permissions to add, shown with a plus sign, which does not occur for other principals. Lateral movement insights identify roles that let a service account in one project impersonate a service account in another, which Google says can create a chain of impersonations across projects. Google provides an organization policy constraint, constraints/iam.automaticIamGrantsForDefaultServiceAccounts, which stops default service accounts in new projects from being automatically granted Editor. Google states that applying it does not remove Editor from existing default service accounts. [The routes page](/routes/) covers what to do when provider tools alone do not finish the job. ## Steps to bring service accounts to least privilege - Enforce the constraint that prevents automatic Editor grants for default service accounts, then list existing default service accounts that still hold roles/editor. - Create a dedicated service account per application, as Google advises, and attach it to the resource instead of the default account. - Remove Editor from default service accounts after granting the specific predefined roles each workload needs. - Use the recent usage view to find service accounts and keys with no use. Disable them, wait through a business cycle, then delete. - List user-managed keys. Replace each with attached service accounts, impersonation or workload identity federation, and disable the key before deleting it. - Review who holds Service Account User and Service Account Token Creator-type access on each service account, and limit it. - Apply IAM recommender suggestions for service accounts, and check lateral movement insights before cross-project permissions are removed. - Repeat on a schedule. See [what least privilege access means](/least-privilege-access/) for the method. ## Roles and credentials to avoid on service accounts | Role or credential | Documented reason (Google, read 16 Nov 2026) | Use instead | | --- | --- | --- | | roles/editor on a default service account | Google says default service accounts get Editor automatically and that sharing one across applications is very risky. | A dedicated service account with predefined roles. | | roles/owner | Google says it allows access to and modification of almost all resources and should be granted only when nearly universal access is required. | A predefined role on the resource. | | User-managed service account keys | Google lists leakage, privilege escalation, exposure of confidential metadata, non-repudiation and malicious credential configuration as threats. | Attached service accounts, impersonation or federation. | | One service account shared across applications | Google says it makes activity hard to trace to the right application. | A service account per application. | | Service Account User granted widely | Google says holders can access everything the service account can. | Grant on the single service account, to specific principals. | | Unused service accounts left enabled | Google recommends identifying and disabling unused service accounts. | Disable, then delete after verification. | ## Measure accounts, keys and the 90-day permission gap Record per project: default service accounts with Editor, user-managed keys by age, service accounts with no recent use, and service accounts that other projects can impersonate. For each active service account, compare permissions in its roles with permissions used in the last 90 days, which is the recommender's window. A key count that falls month on month is the clearest sign of progress, because every removed key removes a standing credential. [Privilege creep](/privilege-creep/) shows why the default Editor grant outlives the project that created it. Related guides: [AWS service identities](/guides/aws-service-identities/) and [Azure service identities](/guides/azure-service-identities/). ## Related pages - [Least privilege for workload identities in Google Cloud](/guides/google-cloud-workload-identities/) - [Least privilege for third-party identities in Google Cloud](/guides/google-cloud-third-party-identities/) - [Least privilege for human users in AWS](/guides/aws-human-identities/) - [All of them](/guides/) ## Sources - Google Cloud IAM: Best practices for using service accounts: https://cloud.google.com/iam/docs/best-practices-service-accounts (read 16 Nov 2026) - Google Cloud IAM: Best practices for managing service account keys: https://cloud.google.com/iam/docs/best-practices-for-managing-service-account-keys (read 16 Nov 2026) - Google Cloud IAM: Service accounts overview: https://cloud.google.com/iam/docs/service-account-overview (read 16 Nov 2026) - Google Cloud IAM: Service agents: https://cloud.google.com/iam/docs/service-agents (read 16 Nov 2026) - Google Cloud IAM: Role recommendations overview: https://cloud.google.com/iam/docs/recommender-overview (read 16 Nov 2026) - Google Cloud IAM: Manage role recommendations: https://cloud.google.com/iam/docs/recommender-managing (read 16 Nov 2026) - Google Cloud: Policy Intelligence overview: https://cloud.google.com/policy-intelligence/docs/overview (read 16 Nov 2026) - Google Cloud IAM: Best practices for using IAM securely: https://cloud.google.com/iam/docs/using-iam-securely (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/guides/google-cloud-service-identities/ --- # Least privilege for workload identities in Google Cloud means attached service accounts and federation with narrow bindings Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. Workload identities in Google Cloud are the service accounts attached to compute and the federated identities of workloads that run outside Google Cloud. Google's documentation, read 16 Nov 2026, describes Workload Identity Federation, direct access and impersonation. This guide covers the tools, steps, roles to avoid and measures. ## Workloads authenticate as attached service accounts or as federated identities Workloads on Google Cloud compute use a service account attached to the resource, so no key is needed. Workloads outside Google Cloud, on AWS, Azure, in GitHub or GitLab pipelines, on Kubernetes clusters or on-premises, use Workload Identity Federation. Google's documentation says federation lets you grant access to on-premises or multicloud workloads using federated identities instead of service account keys, and describes keys as powerful credentials that can present a security risk if not managed correctly. There are two ways to grant access once the identity is federated. Direct access grants IAM roles to the federated principal itself, with principal:// bindings for one subject and principalSet:// bindings for a group or an attribute. Service account impersonation grants the federated identity permission to act as a service account. Google recommends creating a new workload identity pool for each non-Google Cloud environment, such as development, staging or production. ## Google documents attribute conditions, narrow bindings and the recommender for this identity type Google's best practices for Workload Identity Federation say to avoid granting the roles/iam.workloadIdentityUser role to all members of a pool, to grant it to specific external identities or identities that match criteria, and to create a dedicated service account for each application. The same page says to use attribute conditions when federating with GitHub or other multi-tenant identity providers, to use immutable attributes in attribute mappings, to use the provider URL as the audience, to use a single provider per pool to avoid subject collisions, and to manage pools in a dedicated project. It also names an organization policy constraint that can deny creation of new pool providers by default. For the service accounts behind the workloads, the IAM recommender and the service account usage tools apply, as in the [service identities guide](/guides/google-cloud-service-identities/). Policy Troubleshooter can explain why a federated principal was allowed or denied a permission. Google's documentation lists a separate feature for managed workload identities on GKE and Compute Engine, and the Kubernetes federation page describes the cluster setup. Both rely on the same bindings. ## Steps to bring workload identities to least privilege - Inventory workloads outside Google Cloud that use service account keys, and the service account each key belongs to. - Create one workload identity pool per environment in a dedicated project and one provider per pool. - Write attribute conditions that admit only the repositories, clusters or accounts you expect, and require the provider URL as audience. - Create a dedicated service account per application and grant it predefined roles at project or resource scope. - Bind roles/iam.workloadIdentityUser to specific subjects, never to the whole pool. - Where direct access is enough, bind roles to the federated principal and skip the service account. - Delete the service account keys that federation replaces, disabling them first. - Review recommender suggestions for each service account after one full release cycle. ## Roles and bindings to avoid on workload identities | Role or binding | Documented reason (Google, read 16 Nov 2026) | Use instead | | --- | --- | --- | | roles/iam.workloadIdentityUser to the whole pool | Google says to avoid it, so a pool can grow without granting impersonation to every new identity. | Bindings to specific subjects or attributes. | | roles/editor on the attached service account | Google says default service accounts hold Editor and that sharing a powerful account is risky. | A dedicated account with predefined roles. | | Service account keys for external workloads | Google says federation removes the burden of keys, which are powerful credentials. | Workload Identity Federation. | | A pool with no attribute condition on GitHub or another multi-tenant provider | Google says to use attribute conditions with multi-tenant providers. | A condition that admits only your repositories or tenants. | | One service account for several applications | Google says to create a dedicated service account for each application. | One service account per application. | ## Measure bindings, pools and permissions used per service account Count workloads that still use keys against workloads that use federation or attached accounts. Count bindings of workloadIdentityUser that reference a whole pool rather than a subject. For each service account, compare the permissions in its roles with permissions used in the last 90 days from the recommender. Audit logs record the federated subject, which is one reason Google advises a unique subject mapping, so each federated identity can be traced to exactly one external identity. [The principle of least privilege](/principle-of-least-privilege/) explains the logic of the per-permission measure. Related guides: [AWS workload identities](/guides/aws-workload-identities/) and [Azure workload identities](/guides/azure-workload-identities/). ## Related pages - [Least privilege for third-party identities in Google Cloud](/guides/google-cloud-third-party-identities/) - [Least privilege for human users in AWS](/guides/aws-human-identities/) - [Least privilege for service identities in AWS](/guides/aws-service-identities/) - [All of them](/guides/) ## Sources - Google Cloud IAM: Workload Identity Federation: https://cloud.google.com/iam/docs/workload-identity-federation (read 16 Nov 2026) - Google Cloud IAM: Best practices for using Workload Identity Federation: https://cloud.google.com/iam/docs/best-practices-for-using-workload-identity-federation (read 16 Nov 2026) - Google Cloud IAM: Best practices for using service accounts: https://cloud.google.com/iam/docs/best-practices-service-accounts (read 16 Nov 2026) - Google Cloud IAM: Best practices for managing service account keys: https://cloud.google.com/iam/docs/best-practices-for-managing-service-account-keys (read 16 Nov 2026) - Google Cloud IAM: Role recommendations overview: https://cloud.google.com/iam/docs/recommender-overview (read 16 Nov 2026) - Google Cloud IAM: Best practices for using IAM securely: https://cloud.google.com/iam/docs/using-iam-securely (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/guides/google-cloud-workload-identities/ --- # Least privilege for third-party identities in Google Cloud means restricted domains, cross-project controls and federation with narrow bindings Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. Third-party identities in Google Cloud are the external users, cross-project service accounts and federated partner identities that reach your resources. Google's documentation, read 16 Nov 2026, covers domain-restricted sharing, cross-project service account use and Workload Identity Federation. This guide gives the controls, steps and measures. ## Outside parties reach Google Cloud as external principals, cross-project service accounts or federated identities Domain-restricted sharing limits which identities can be granted IAM roles in your organization. Google says that when it is active, only principals that belong to allowed domains or organizations can be granted roles, and that for organizations created on or after 3 May 2024 the iam.allowedPolicyMemberDomains constraint is enforced by default with your domain as the only allowed value. Service accounts cross project boundaries in two ways. A service account in one project can be attached to a resource in another project, which Google gates with the iam.disableCrossProjectServiceAccountUsage constraint, enforced by default at project level. A service account can also be granted permission to impersonate service accounts elsewhere. Federation covers partners and external systems that already have an identity provider. Google says Workload Identity Federation can grant access to federated identities from AWS, Azure, GitHub, GitLab, OIDC and SAML providers and others, without service account keys. ## Google documents lateral movement insights, attribute conditions and narrow impersonation bindings Google describes lateral movement as a service account in one project having permission to impersonate a service account in another, which can chain across projects. IAM recommender provides lateral movement insights that identify those roles, and Google states that attaching a service account across projects is not lateral movement because it grants no impersonation permission. For federated partners, Google's best practices say to use attribute conditions when federating with multi-tenant identity providers, to avoid granting roles/iam.workloadIdentityUser to all pool members, to use a dedicated service account for each application and to use a unique google.subject mapping so each external identity can be traced in audit logs. Google also advises a single provider per pool, one pool per environment, and the pool provider URL as the audience. A partner that needs a separate relationship should get a separate pool, so removing the partner is one deletion. When enabling cross-project service account use, Google discourages undoing the change in production and warns that resources might stop working, so plan the exit before the entry. Read [the first 90 days](/first-90-days/) for ordering this against other work. ## Steps to bring third-party access in Google Cloud to least privilege - List allow policy members from outside your domains, using Policy Analyzer, and the owner of each relationship. - Review the domain-restricted sharing setting and keep your own domain as the only allowed value unless a named partner domain is required. - List projects where cross-project service account use is enabled and the service accounts used across projects. - Open lateral movement insights and remove impersonation roles that cross projects without a reason. - Give each partner its own workload identity pool, with an attribute condition that admits only the partner's expected identities. - Bind partner identities to a dedicated service account with predefined roles, or grant roles directly to the federated principal. - Replace any service account key given to a partner with federation. - Set an end date for each relationship and remove the pool or binding on that date. ## Settings and roles to avoid for third parties | Setting or role | Documented reason (Google, read 16 Nov 2026) | Use instead | | --- | --- | --- | | Allowing arbitrary domains in sharing policy | Google says only principals in allowed domains can be granted roles when the restriction is active. | Your domain plus named partner domains. | | Service account keys issued to a partner | Google says keys are powerful credentials and federation removes their burden. | Workload Identity Federation. | | roles/iam.workloadIdentityUser to all pool members | Google says to avoid it and bind specific external identities. | Subject or attribute bindings. | | Federating a multi-tenant provider without an attribute condition | Google says to use attribute conditions with GitHub and other multi-tenant providers. | A condition naming your tenant or repository. | | Cross-project impersonation roles with no owner | Google says they can chain impersonation across projects. | Remove after reviewing lateral movement insights. | | roles/editor on a shared service account | Google says default service accounts hold Editor and sharing them is very risky. | A dedicated account with predefined roles. | ## Measure external principals, cross-project grants and used permissions Count external principals with roles, the share with a named owner and an end date, and the count of service accounts with cross-project impersonation. For each service account that serves a partner, compare the permissions in its roles with those used in the last 90 days. A partner that appears in the allow policy but has no recorded activity is access nobody uses. [Privilege creep](/privilege-creep/) explains how this builds over contracts, and [the glossary](/glossary/) defines the terms. Related guides: [AWS third-party identities](/guides/aws-third-party-identities/) and [Azure third-party identities](/guides/azure-third-party-identities/). ## Related pages - [Least privilege for human users in AWS](/guides/aws-human-identities/) - [Least privilege for service identities in AWS](/guides/aws-service-identities/) - [Least privilege for workload identities in AWS](/guides/aws-workload-identities/) - [All of them](/guides/) ## Sources - Google Cloud: Restrict identities with domain-restricted sharing: https://cloud.google.com/resource-manager/docs/organization-policy/restricting-domains (read 16 Nov 2026) - Google Cloud IAM: Attach service accounts to resources: https://cloud.google.com/iam/docs/attach-service-accounts (read 16 Nov 2026) - Google Cloud IAM: Best practices for using service accounts: https://cloud.google.com/iam/docs/best-practices-service-accounts (read 16 Nov 2026) - Google Cloud IAM: Workload Identity Federation: https://cloud.google.com/iam/docs/workload-identity-federation (read 16 Nov 2026) - Google Cloud IAM: Best practices for using Workload Identity Federation: https://cloud.google.com/iam/docs/best-practices-for-using-workload-identity-federation (read 16 Nov 2026) - Google Cloud IAM: Role recommendations overview: https://cloud.google.com/iam/docs/recommender-overview (read 16 Nov 2026) - Google Cloud IAM: Best practices for managing service account keys: https://cloud.google.com/iam/docs/best-practices-for-managing-service-account-keys (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/guides/google-cloud-third-party-identities/ --- # 40 terms, each with a sourced example Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. Plain definitions of the terms used in cloud least privilege, from permissions boundary and service control policy to privilege creep and standing access. Each definition stands alone in one sentence, is tied to a provider or NIST page that was read, and shows the date it was read. - [Access debt](/glossary/access-debt/): Access debt is the accumulated gap between the access people and workloads hold in a cloud account and the access they need. - [Access key rotation](/glossary/access-key-rotation/): Access key rotation is replacing an IAM user access key with a new one and deleting the old one on a schedule. - [Access level](/glossary/access-level/): An access level is an AWS classification of an action as List, Read, Write, Permissions management or Tagging. - [Active assignment](/glossary/active-assignment/): An active assignment is a Microsoft Entra Privileged Identity Management assignment that gives the user the role's privileges without any activation step. - [Assignable scope](/glossary/assignable-scope/): An assignable scope is a field in an Azure role definition that lists where the role can be assigned, such as subscriptions or management groups. - [Attribute-based access control](/glossary/attribute-based-access-control/): Attribute-based access control grants or denies access by comparing attributes of the caller, the resource and the request, such as tags. - [Basic role](/glossary/basic-role/): A basic role is a Google Cloud role that provides broad access across services. - [Built-in role](/glossary/built-in-role/): A built-in role is an Azure role defined by Microsoft that can be assigned without writing a role definition. - [Custom policy check](/glossary/custom-policy-check/): A custom policy check is an IAM Access Analyzer test that compares a policy update against your security standard, for example to see if it grants new access compared with the existing version. - [Custom role](/glossary/custom-role/): A custom role is a role whose permission list the customer writes, used when no provider-supplied role matches the task. - [Data action](/glossary/data-action/): A data action is an Azure permission for operations on the data inside a resource, such as reading a blob, as opposed to managing the resource itself. - [Data perimeter](/glossary/data-perimeter/): A data perimeter is a set of always-on permissions guardrails around identities, resources and networks that helps prevent unintended access. - [Deny policy](/glossary/deny-policy/): A deny policy is a Google Cloud policy of deny rules that stop chosen principals from using chosen permissions, whatever roles they hold. - [Eligible assignment](/glossary/eligible-assignment/): An eligible assignment is a Microsoft Entra Privileged Identity Management assignment that requires the user to activate the role before using it. - [IAM Conditions](/glossary/iam-conditions/): IAM Conditions are Google Cloud rules that make a role binding, deny rule or boundary apply only when an attribute test passes. - [Job function policy](/glossary/job-function-policy/): A job function policy is an AWS managed policy that grants the permissions typical for a role in an organization, such as database administrator or security auditor. - [Last accessed information](/glossary/last-accessed-information/): Last accessed information is a record of when an IAM identity or policy last used a service or action. - [Long-term credentials](/glossary/long-term-credentials/): Long-term credentials are keys that stay valid until someone deletes or deactivates them, such as AWS IAM user access keys. - [Managed policy](/glossary/managed-policy/): A managed policy is a standalone IAM policy that exists as its own object and can be attached to many users, groups and roles. - [MFA condition](/glossary/mfa-condition/): An MFA condition is a policy condition that allows an action only when the caller authenticated with multi-factor authentication. - [Permission set](/glossary/permission-set/): A permission set is a template in AWS IAM Identity Center that defines a collection of one or more IAM policies. - [Policy generation](/glossary/policy-generation/): Policy generation is the creation of an IAM policy from observed activity. - [Policy inheritance](/glossary/policy-inheritance/): Policy inheritance is the Google Cloud rule that a resource receives the allow policies of its parent. - [Policy validation](/glossary/policy-validation/): Policy validation is the check of an IAM policy against policy grammar and best practices before it is saved. - [Predefined role](/glossary/predefined-role/): A predefined role is a Google Cloud role that provides granular access to one service and is created and maintained by Google. - [Principal access boundary](/glossary/principal-access-boundary/): A principal access boundary is a Google Cloud policy that sets which resources a group of principals is eligible to access. - [Privileged role](/glossary/privileged-role/): A privileged role is a built-in Azure role whose permissions include broad management or the ability to assign access to others. - [Resource scope](/glossary/resource-scope/): Resource scope is the set of resources a role assignment applies to. - [Right-sizing](/glossary/right-sizing/): Right-sizing is reducing an identity's permissions to the set it actually uses. - [Role assignment](/glossary/role-assignment/): A role assignment attaches a role definition to a principal at a scope in Azure. - [Role assignment condition](/glossary/role-assignment-condition/): A role assignment condition is an extra check added to an Azure role assignment that tests attributes of the request, giving attribute-based access control on top of RBAC. - [Role chaining](/glossary/role-chaining/): Role chaining is using one AWS role to assume a second role. - [Role definition](/glossary/role-definition/): A role definition is the Azure object that lists the permissions a role allows. - [Role recommendation](/glossary/role-recommendation/): A role recommendation is a Google Cloud suggestion from the IAM recommender to remove or replace a role that gives a principal excess permissions. - [Role-based access control](/glossary/role-based-access-control/): Role-based access control is an authorization approach in which permissions are grouped into roles and people or workloads receive roles instead of individual permissions. - [Service account impersonation](/glossary/service-account-impersonation/): Service account impersonation is a Google Cloud principal authenticating as a service account to gain its permissions. - [Session policy](/glossary/session-policy/): A session policy is an AWS policy passed when a role is assumed or a federated user session is created. - [Temporary credentials](/glossary/temporary-credentials/): Temporary credentials are short-lived keys issued by AWS Security Token Service that expire on their own, and they are the basis of IAM roles and federation. - [Unused access finding](/glossary/unused-access-finding/): An unused access finding is an IAM Access Analyzer result that reports access granted but not used during a chosen usage window. - [Zone of trust](/glossary/zone-of-trust/): A zone of trust is the account or organization that an IAM Access Analyzer external access analyzer treats as safe. ## Related pages - [News](/news/) - [Least privilege for 20 cloud services](/services/) - [Routes to least privilege: tools and platforms](/routes/) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/glossary/ --- # Access debt Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. Access debt is the accumulated gap between the access people and workloads hold in a cloud account and the access they need. It builds up as roles are added, shared and left in place, and it is cleared by removing access, not by recording it. ## An example, sourced Act's platform page says its product helps security teams "clear access debt, prevent drift, and enforce the boundaries that protect crown jewels", and says it eliminates unmanaged access at the infrastructure layer (read 16 Nov 2026). ## Why it matters here The term turns a vague worry into a backlog that can be sized and worked down. A team that counts the access it can remove can say when the backlog shrinks. ## Related pages - [Right-sizing: meaning and an example](/glossary/right-sizing/) - [Role-based access control: meaning and an example](/glossary/role-based-access-control/) - [Attribute-based access control: meaning and an example](/glossary/attribute-based-access-control/) - [All of them](/glossary/) ## Sources - https://act.security/platform (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/glossary/access-debt/ --- # Right-sizing Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. Right-sizing is reducing an identity's permissions to the set it actually uses. It starts from observed activity, proposes a smaller grant, and ends when the smaller grant replaces the larger one. Vendors and providers use the word for both the analysis and the change. ## An example, sourced Microsoft's overview of Entra Permissions Management (a product Microsoft has since retired) said it "detects, automatically right-sizes, and continuously monitors unused and excessive permissions" across clouds. ## Why it matters here The word hides a question about who applies the change. A tool that only recommends a smaller grant leaves the work to the account owner. ## Related pages - [Role-based access control: meaning and an example](/glossary/role-based-access-control/) - [Attribute-based access control: meaning and an example](/glossary/attribute-based-access-control/) - [Permission set: meaning and an example](/glossary/permission-set/) - [All of them](/glossary/) ## Sources - https://learn.microsoft.com/en-us/entra/permissions-management/overview (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/glossary/right-sizing/ --- # Role-based access control Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. Role-based access control is an authorization approach in which permissions are grouped into roles and people or workloads receive roles instead of individual permissions. The scope of the role assignment decides which resources the permissions reach. The model works best when each role is small and each assignment is limited to the scope that task needs. ## An example, sourced Microsoft's Azure documentation says Azure RBAC helps you manage who has access to Azure resources, what they can do with those resources, and what areas they have access to. ## Why it matters here Roles are only as narrow as their permission lists. A broad role granted widely gives little protection, so the role contents matter more than the model. ## Related pages - [Attribute-based access control: meaning and an example](/glossary/attribute-based-access-control/) - [Permission set: meaning and an example](/glossary/permission-set/) - [Custom role: meaning and an example](/glossary/custom-role/) - [All of them](/glossary/) ## Sources - https://learn.microsoft.com/en-us/azure/role-based-access-control/overview (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/glossary/role-based-access-control/ --- # Attribute-based access control Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. Attribute-based access control grants or denies access by comparing attributes of the caller, the resource and the request, such as tags. In AWS, the attributes are tags on IAM principals and resources, and one policy can serve many projects. It pairs naturally with role sessions, since the attributes can come from the identity provider. ## An example, sourced AWS documentation describes three IAM roles with the access-project tag key and a policy that allows access when the principal's tag matches the tag on the resource, so new resources are covered without policy edits. ## Why it matters here Tags let one small policy replace many project-specific policies, which keeps each grant narrow. Tag control then becomes part of the access boundary. ## Related pages - [Permission set: meaning and an example](/glossary/permission-set/) - [Custom role: meaning and an example](/glossary/custom-role/) - [Predefined role: meaning and an example](/glossary/predefined-role/) - [All of them](/glossary/) ## Sources - https://docs.aws.amazon.com/IAM/latest/UserGuide/introduction_attribute-based-access-control.html (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/glossary/attribute-based-access-control/ --- # Permission set Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. A permission set is a template in AWS IAM Identity Center that defines a collection of one or more IAM policies. Administrators assign it to users and groups for chosen AWS accounts, and Identity Center creates the matching role in each account. ## An example, sourced AWS documentation gives a Database Admin permission set that includes policies for RDS, DynamoDB and Aurora, assigned once to a list of target accounts for database administrators. ## Why it matters here One permission set reaches many accounts, so an overly broad one spreads far. Keep each set to one job function. ## Related pages - [Custom role: meaning and an example](/glossary/custom-role/) - [Predefined role: meaning and an example](/glossary/predefined-role/) - [Basic role: meaning and an example](/glossary/basic-role/) - [All of them](/glossary/) ## Sources - https://docs.aws.amazon.com/singlesignon/latest/userguide/permissionsetsconcept.html (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/glossary/permission-set/ --- # Custom role Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. A custom role is a role whose permission list the customer writes, used when no provider-supplied role matches the task. In Google Cloud it is a user-specified list of permissions; in Azure it is a role definition you create with chosen actions and assignable scopes. ## An example, sourced Microsoft's documentation shows a custom role described as "Can monitor and restart virtual machines" with assignable scopes naming two subscriptions and one management group. ## Why it matters here A custom role is the main way to land below a broad built-in role. It also needs upkeep, because new permissions are not added to it automatically. ## Related pages - [Predefined role: meaning and an example](/glossary/predefined-role/) - [Basic role: meaning and an example](/glossary/basic-role/) - [Managed policy: meaning and an example](/glossary/managed-policy/) - [All of them](/glossary/) ## Sources - https://learn.microsoft.com/en-us/azure/role-based-access-control/custom-roles (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/glossary/custom-role/ --- # Predefined role Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. A predefined role is a Google Cloud role that provides granular access to one service and is created and maintained by Google. Customers pick it instead of writing their own permission list, and Google updates its permissions as the service changes. ## An example, sourced Google's roles overview says predefined roles provide granular access for a specific service and are managed by Google Cloud, and that the role ID is everything after roles/ in the role name. ## Why it matters here Predefined roles are narrower than basic roles and cheaper to run than custom roles. They are the usual first choice when tightening a grant. ## Related pages - [Basic role: meaning and an example](/glossary/basic-role/) - [Managed policy: meaning and an example](/glossary/managed-policy/) - [Last accessed information: meaning and an example](/glossary/last-accessed-information/) - [All of them](/glossary/) ## Sources - https://cloud.google.com/iam/docs/roles-overview (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/glossary/predefined-role/ --- # Basic role Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. A basic role is a Google Cloud role that provides broad access across services. The legacy basic roles are Owner, Editor and Viewer, and newer basic roles named Admin, Writer and Reader have been added. They are rarely the narrowest choice for a given job. ## An example, sourced Google's documentation says basic roles include thousands of permissions across all Google Cloud services, and lists roles/viewer, roles/writer and roles/reader among them. ## Why it matters here Basic roles carry the widest grant in a project, so they are the first place to look when reducing access. ## Related pages - [Managed policy: meaning and an example](/glossary/managed-policy/) - [Last accessed information: meaning and an example](/glossary/last-accessed-information/) - [Unused access finding: meaning and an example](/glossary/unused-access-finding/) - [All of them](/glossary/) ## Sources - https://cloud.google.com/iam/docs/roles-overview (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/glossary/basic-role/ --- # Managed policy Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. A managed policy is a standalone IAM policy that exists as its own object and can be attached to many users, groups and roles. AWS creates and updates the AWS managed ones, and customers write the customer managed ones. Both kinds can be attached to many identities at once. ## An example, sourced AWS documentation names arn:aws:iam::aws:policy/IAMReadOnlyAccess as an AWS managed policy, and says ReadOnlyAccess is updated when AWS launches a new service. ## Why it matters here An AWS managed policy can change under you when AWS adds permissions. A customer managed policy changes only when you change it, which makes reviews stable. ## Related pages - [Last accessed information: meaning and an example](/glossary/last-accessed-information/) - [Unused access finding: meaning and an example](/glossary/unused-access-finding/) - [Role recommendation: meaning and an example](/glossary/role-recommendation/) - [All of them](/glossary/) ## Sources - https://docs.aws.amazon.com/IAM/latest/UserGuide/access_policies_managed-vs-inline.html (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/glossary/managed-policy/ --- # Last accessed information Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. Last accessed information is a record of when an IAM identity or policy last used a service or action. It tells an administrator which permissions have gone unused, so they can be removed without guessing. AWS also tracks it for organizations, so an administrator can review a whole account set. ## An example, sourced AWS documentation says IAM provides last accessed information to identify unused permissions, and that you can view it for identities or policies in IAM or AWS Organizations, including allowed service and allowed action information. ## Why it matters here It is the evidence behind a safe removal. Without it, an administrator cannot tell an unused permission from one used once a year. ## Related pages - [Unused access finding: meaning and an example](/glossary/unused-access-finding/) - [Role recommendation: meaning and an example](/glossary/role-recommendation/) - [Access level: meaning and an example](/glossary/access-level/) - [All of them](/glossary/) ## Sources - https://docs.aws.amazon.com/IAM/latest/UserGuide/access_policies_last-accessed.html (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/glossary/last-accessed-information/ --- # Unused access finding Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. An unused access finding is an IAM Access Analyzer result that reports access granted but not used during a chosen usage window. The finding types include unused roles, unused access keys and passwords, and unused permissions. Each finding names the identity and the access that went unused, so an owner can act on it. ## An example, sourced AWS documentation says an unused access analyzer continuously monitors all IAM roles and users in the organization and accounts, and lists unused roles and unused IAM user access keys and passwords among the finding types. ## Why it matters here Findings turn a long permission list into a short work queue. AWS charges for unused access analysis, while external access findings are free. ## Related pages - [Role recommendation: meaning and an example](/glossary/role-recommendation/) - [Access level: meaning and an example](/glossary/access-level/) - [Resource scope: meaning and an example](/glossary/resource-scope/) - [All of them](/glossary/) ## Sources - https://docs.aws.amazon.com/IAM/latest/UserGuide/access-analyzer-findings.html (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/glossary/unused-access-finding/ --- # Role recommendation Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. A role recommendation is a Google Cloud suggestion from the IAM recommender to remove or replace a role that gives a principal excess permissions. A person reviews it and decides whether to apply it. Google presents it in the console and through the API for review. ## An example, sourced Google's documentation says the maximum observation period for role recommendations is 90 days, and that recommendations suggest removing or replacing a role that grants excess permissions. ## Why it matters here The 90 day window matters. A permission used only in a quarterly job may fall outside it, so check the workload before applying the change. ## Related pages - [Access level: meaning and an example](/glossary/access-level/) - [Resource scope: meaning and an example](/glossary/resource-scope/) - [Session policy: meaning and an example](/glossary/session-policy/) - [All of them](/glossary/) ## Sources - https://cloud.google.com/iam/docs/recommender-overview (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/glossary/role-recommendation/ --- # Access level Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. An access level is an AWS classification of an action as List, Read, Write, Permissions management or Tagging. Policy summaries use it to show how much of each kind a policy allows, so a reviewer can spot write or permission actions fast. ## An example, sourced AWS documentation describes policy summaries that report access to "at least one but not all" Amazon EC2 Write actions, grouped by service and access level. ## Why it matters here Two policies with the same service list can differ sharply in risk. Permissions management actions deserve extra scrutiny because they change who else has access. ## Related pages - [Resource scope: meaning and an example](/glossary/resource-scope/) - [Session policy: meaning and an example](/glossary/session-policy/) - [Temporary credentials: meaning and an example](/glossary/temporary-credentials/) - [All of them](/glossary/) ## Sources - https://docs.aws.amazon.com/IAM/latest/UserGuide/access_policies_understand-policy-summary-access-level-summaries.html (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/glossary/access-level/ --- # Resource scope Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. Resource scope is the set of resources a role assignment applies to. In Azure the levels are management group, subscription, resource group and resource, and each level down narrows the reach. It is the second half of every Azure grant, next to the role itself. ## An example, sourced Microsoft's documentation says that by limiting the scope you limit what resources are at risk if the security principal is ever compromised, and lists four scope levels in a parent-child structure. ## Why it matters here The same role at resource group scope reaches far fewer things than at subscription scope. Narrowing scope is often the cheapest tightening step. ## Related pages - [Session policy: meaning and an example](/glossary/session-policy/) - [Temporary credentials: meaning and an example](/glossary/temporary-credentials/) - [Access key rotation: meaning and an example](/glossary/access-key-rotation/) - [All of them](/glossary/) ## Sources - https://learn.microsoft.com/en-us/azure/role-based-access-control/scope-overview (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/glossary/resource-scope/ --- # Session policy Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. A session policy is an AWS policy passed when a role is assumed or a federated user session is created. It limits the permissions the session can use but never adds any. It works as a per-session ceiling applied on top of the role's own policies. ## An example, sourced AWS documentation says session policies limit the permissions that the role or user's identity-based policies grant to the session, are passed with the AWS CLI or API, and do not grant permissions themselves. ## Why it matters here A session policy lets one role serve many callers while each session receives only a slice. The role stays broad, and the session does not. ## Related pages - [Temporary credentials: meaning and an example](/glossary/temporary-credentials/) - [Access key rotation: meaning and an example](/glossary/access-key-rotation/) - [MFA condition: meaning and an example](/glossary/mfa-condition/) - [All of them](/glossary/) ## Sources - https://docs.aws.amazon.com/IAM/latest/UserGuide/access_policies.html (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/glossary/session-policy/ --- # Temporary credentials Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. Temporary credentials are short-lived keys issued by AWS Security Token Service that expire on their own, and they are the basis of IAM roles and federation. Because they expire, nobody has to revoke or rotate them. AWS says they can be requested again as long as the caller still has permission. ## An example, sourced AWS documentation says temporary credentials can last from a few minutes to several hours, are generated dynamically when requested, and let you avoid embedding long-term AWS credentials in an application. ## Why it matters here A stolen temporary credential stops working at expiry. Replacing long-term keys with them cuts the time a leaked secret is useful. ## Related pages - [Access key rotation: meaning and an example](/glossary/access-key-rotation/) - [MFA condition: meaning and an example](/glossary/mfa-condition/) - [Data perimeter: meaning and an example](/glossary/data-perimeter/) - [All of them](/glossary/) ## Sources - https://docs.aws.amazon.com/IAM/latest/UserGuide/id_credentials_temp.html (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/glossary/temporary-credentials/ --- # Access key rotation Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. Access key rotation is replacing an IAM user access key with a new one and deleting the old one on a schedule. It limits how long a leaked key stays valid, and AWS advises avoiding long-term keys where roles can be used instead. ## An example, sourced AWS documentation says an IAM user can have a maximum of two access keys, which allows a second key to be created before the first is deleted, and says to use temporary credentials instead of access keys as a best practice. ## Why it matters here Rotation shortens exposure but does not narrow what a key can do. Removing the key in favor of a role removes the problem. ## Related pages - [MFA condition: meaning and an example](/glossary/mfa-condition/) - [Data perimeter: meaning and an example](/glossary/data-perimeter/) - [Built-in role: meaning and an example](/glossary/built-in-role/) - [All of them](/glossary/) ## Sources - https://docs.aws.amazon.com/IAM/latest/UserGuide/id_credentials_access-keys.html (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/glossary/access-key-rotation/ --- # MFA condition Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. An MFA condition is a policy condition that allows an action only when the caller authenticated with multi-factor authentication. It narrows a grant to sessions that passed a second factor. It can sit in an identity policy, a resource policy or a role trust policy. ## An example, sourced AWS documentation shows a trust policy for sts:AssumeRole with the condition "Bool": {"aws:MultiFactorAuthPresent": "true"}, and notes the key is present only when the user authenticates with short-term credentials. ## Why it matters here It protects sensitive actions without adding roles. The check is only as good as the factor behind it. ## Related pages - [Data perimeter: meaning and an example](/glossary/data-perimeter/) - [Built-in role: meaning and an example](/glossary/built-in-role/) - [Eligible assignment: meaning and an example](/glossary/eligible-assignment/) - [All of them](/glossary/) ## Sources - https://docs.aws.amazon.com/IAM/latest/UserGuide/id_credentials_mfa_configure-api-require.html (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/glossary/mfa-condition/ --- # Data perimeter Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. A data perimeter is a set of always-on permissions guardrails around identities, resources and networks that helps prevent unintended access. In AWS it is built from policies that fix who may reach which data from where. The perimeter works as a second layer, because it holds even when a role is broader than intended. ## An example, sourced The AWS whitepaper Building a Data Perimeter on AWS, dated June 13, 2023, describes perimeter controls that protect against unintended access and configuration errors for identities, resources and networks. ## Why it matters here A perimeter caps the damage when an individual policy is too loose. It complements least privilege on each identity. ## Related pages - [Built-in role: meaning and an example](/glossary/built-in-role/) - [Eligible assignment: meaning and an example](/glossary/eligible-assignment/) - [Active assignment: meaning and an example](/glossary/active-assignment/) - [All of them](/glossary/) ## Sources - https://docs.aws.amazon.com/whitepapers/latest/building-a-data-perimeter-on-aws/building-a-data-perimeter-on-aws.html (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/glossary/data-perimeter/ --- # Built-in role Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. A built-in role is an Azure role defined by Microsoft that can be assigned without writing a role definition. Each has a name, a description and an ID, and Microsoft maintains the permission list. Roles of this kind range from narrow data readers to Owner. ## An example, sourced Microsoft's built-in roles page lists Storage Blob Data Reader, ID b7e6dc6d-f1e8-4753-8033-0f276bb0955b, described as "Read and list Azure Storage containers and blobs". ## Why it matters here Choosing a narrow built-in role such as a data reader avoids both Contributor and a custom role. Check the role's actions before assigning it. ## Related pages - [Eligible assignment: meaning and an example](/glossary/eligible-assignment/) - [Active assignment: meaning and an example](/glossary/active-assignment/) - [Role assignment: meaning and an example](/glossary/role-assignment/) - [All of them](/glossary/) ## Sources - https://learn.microsoft.com/en-us/azure/role-based-access-control/built-in-roles (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/glossary/built-in-role/ --- # Eligible assignment Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. An eligible assignment is a Microsoft Entra Privileged Identity Management assignment that requires the user to activate the role before using it. The user holds the right to take the role, not the role itself. Activation can require a justification, approval or a multifactor check, and the role then applies for a limited time. ## An example, sourced Microsoft's documentation says an eligible assignment requires a user to perform one or more actions to use the role, and that activation might need an MFA check, a business justification or approval. ## Why it matters here It removes standing privilege without removing the person's ability to do the job. The access is the same once activated. ## Related pages - [Active assignment: meaning and an example](/glossary/active-assignment/) - [Role assignment: meaning and an example](/glossary/role-assignment/) - [Role definition: meaning and an example](/glossary/role-definition/) - [All of them](/glossary/) ## Sources - https://learn.microsoft.com/en-us/entra/id-governance/privileged-identity-management/pim-configure (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/glossary/eligible-assignment/ --- # Active assignment Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. An active assignment is a Microsoft Entra Privileged Identity Management assignment that gives the user the role's privileges without any activation step. It is the opposite of an eligible assignment. Anyone holding it can use the privileges at any time, which is why it counts as standing access. ## An example, sourced Microsoft's documentation says an active assignment does not require a user to perform any action to use the role, so users assigned as active have the privileges of the role. ## Why it matters here Counting active assignments to privileged roles shows how much standing access exists. Each one that can be made eligible is a candidate for removal. ## Related pages - [Role assignment: meaning and an example](/glossary/role-assignment/) - [Role definition: meaning and an example](/glossary/role-definition/) - [IAM Conditions: meaning and an example](/glossary/iam-conditions/) - [All of them](/glossary/) ## Sources - https://learn.microsoft.com/en-us/entra/id-governance/privileged-identity-management/pim-configure (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/glossary/active-assignment/ --- # Role assignment Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. A role assignment attaches a role definition to a principal at a scope in Azure. It decides who gets the permissions, where they apply, and, with a condition, under what circumstances. One person can hold several assignments at different scopes, and the assignments combine. ## An example, sourced Microsoft's documentation says a role assignment is associated with a role definition, and lists properties including the scope and the roleDefinitionId, the unique ID of the role. ## Why it matters here Least privilege in Azure is mostly decided at assignment time, by choosing the role, the principal and the scope together. ## Related pages - [Role definition: meaning and an example](/glossary/role-definition/) - [IAM Conditions: meaning and an example](/glossary/iam-conditions/) - [Principal access boundary: meaning and an example](/glossary/principal-access-boundary/) - [All of them](/glossary/) ## Sources - https://learn.microsoft.com/en-us/azure/role-based-access-control/role-assignments (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/glossary/role-assignment/ --- # Role definition Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. A role definition is the Azure object that lists the permissions a role allows. It names actions, data actions, exclusions and the scopes where the role can be assigned. Custom roles are written as role definitions, and built-in roles are published in the same form. ## An example, sourced Microsoft's documentation lists the properties Actions, NotActions, DataActions, NotDataActions, AssignableScopes and Condition in a role definition, with dataActions covering operations on data in an object. ## Why it matters here Reading the definition is the only reliable way to know what a role permits. Names such as Contributor can hide wide reach. ## Related pages - [IAM Conditions: meaning and an example](/glossary/iam-conditions/) - [Principal access boundary: meaning and an example](/glossary/principal-access-boundary/) - [Deny policy: meaning and an example](/glossary/deny-policy/) - [All of them](/glossary/) ## Sources - https://learn.microsoft.com/en-us/azure/role-based-access-control/role-definitions (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/glossary/role-definition/ --- # IAM Conditions Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. IAM Conditions are Google Cloud rules that make a role binding, deny rule or boundary apply only when an attribute test passes. They support attribute-based access control on Google Cloud resources. The conditions are written in an expression language and checked at the time of each request. ## An example, sourced Google's documentation says conditions can be used in allow policy role bindings, deny policy rules and Principal Access Boundary bindings, with attributes such as request.time and resource.name, and cannot be used with the legacy Owner, Editor and Viewer roles. ## Why it matters here A condition narrows a grant by time, resource name or tag without making a new role. It is a useful tool when a predefined role is close but too wide. ## Related pages - [Principal access boundary: meaning and an example](/glossary/principal-access-boundary/) - [Deny policy: meaning and an example](/glossary/deny-policy/) - [Policy validation: meaning and an example](/glossary/policy-validation/) - [All of them](/glossary/) ## Sources - https://cloud.google.com/iam/docs/conditions-overview (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/glossary/iam-conditions/ --- # Principal access boundary Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. A principal access boundary is a Google Cloud policy that sets which resources a group of principals is eligible to access. It limits what allow policies can grant to those principals. It is a ceiling set by the organization, and it does not grant access on its own. ## An example, sourced Google's documentation says a rule defines a set of resources that principals are eligible to access, and that a permission a policy cannot block stays usable on any resource regardless of the policies the principal is subject to. ## Why it matters here A boundary caps a principal even if a role is granted by mistake. Check which permissions the enforcement version can block. ## Related pages - [Deny policy: meaning and an example](/glossary/deny-policy/) - [Policy validation: meaning and an example](/glossary/policy-validation/) - [Custom policy check: meaning and an example](/glossary/custom-policy-check/) - [All of them](/glossary/) ## Sources - https://cloud.google.com/iam/docs/principal-access-boundary-policies (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/glossary/principal-access-boundary/ --- # Deny policy Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. A deny policy is a Google Cloud policy of deny rules that stop chosen principals from using chosen permissions, whatever roles they hold. It sets a hard limit above allow policies. Deny rules are evaluated before allow policies, so a deny wins when both apply. ## An example, sourced Google's documentation says deny rules prevent certain principals from using certain permissions regardless of the roles granted, and gives iam.googleapis.com/roles.delete as an example permission name. ## Why it matters here Deny rules let an organization fence off dangerous permissions once, instead of removing them from every role. ## Related pages - [Policy validation: meaning and an example](/glossary/policy-validation/) - [Custom policy check: meaning and an example](/glossary/custom-policy-check/) - [Service account impersonation: meaning and an example](/glossary/service-account-impersonation/) - [All of them](/glossary/) ## Sources - https://cloud.google.com/iam/docs/deny-overview (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/glossary/deny-policy/ --- # Policy validation Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. Policy validation is the check of an IAM policy against policy grammar and best practices before it is saved. IAM Access Analyzer reports security warnings, errors, general warnings and suggestions. The same checks run through the console, the AWS CLI and the AWS API. ## An example, sourced AWS documentation says Access Analyzer validates a policy against IAM policy grammar and AWS best practices, shows findings when a managed policy is created or edited in the console, and gives no findings for inline group policies. ## Why it matters here Validation catches overbroad or malformed grants before they reach an account, where removal is harder. ## Related pages - [Custom policy check: meaning and an example](/glossary/custom-policy-check/) - [Service account impersonation: meaning and an example](/glossary/service-account-impersonation/) - [Role chaining: meaning and an example](/glossary/role-chaining/) - [All of them](/glossary/) ## Sources - https://docs.aws.amazon.com/IAM/latest/UserGuide/access-analyzer-policy-validation.html (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/glossary/policy-validation/ --- # Custom policy check Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. A custom policy check is an IAM Access Analyzer test that compares a policy update against your security standard, for example to see if it grants new access compared with the existing version. AWS charges for each check. The check complements policy validation, which tests grammar and best practices rather than access change. ## An example, sourced AWS documentation says you can check whether an updated policy grants new access compared to the existing version while editing in the JSON policy editor, and that a charge applies to each check for new access. ## Why it matters here A check for new access can run in a change pipeline, so a pull request that widens a policy is flagged. ## Related pages - [Service account impersonation: meaning and an example](/glossary/service-account-impersonation/) - [Role chaining: meaning and an example](/glossary/role-chaining/) - [Privileged role: meaning and an example](/glossary/privileged-role/) - [All of them](/glossary/) ## Sources - https://docs.aws.amazon.com/IAM/latest/UserGuide/access-analyzer-custom-policy-checks.html (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/glossary/custom-policy-check/ --- # Service account impersonation Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. Service account impersonation is a Google Cloud principal authenticating as a service account to gain its permissions. It lets a caller act with the service account's access without holding a key. Only principals holding the right permissions can do it, and it avoids creating a service account key. ## An example, sourced Google's documentation says impersonation lets an authenticated principal access whatever the service account can access, and names the Service Account Token Creator role (roles/iam.serviceAccountTokenCreator) as one that holds the needed permission. ## Why it matters here Whoever can impersonate an account inherits its access. Granting the token creator role is as sensitive as granting the account's own roles. ## Related pages - [Role chaining: meaning and an example](/glossary/role-chaining/) - [Privileged role: meaning and an example](/glossary/privileged-role/) - [Policy generation: meaning and an example](/glossary/policy-generation/) - [All of them](/glossary/) ## Sources - https://cloud.google.com/iam/docs/service-account-impersonation (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/glossary/service-account-impersonation/ --- # Role chaining Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. Role chaining is using one AWS role to assume a second role. The second role's access depends on the first role's permission to assume it, so trust has to be read along the whole chain. Each hop in the chain gives the caller a fresh set of short-term credentials. ## An example, sourced AWS documentation says role chaining is when you use a role to assume a second role, and gives the example of RoleA's short-term credentials enabling User1 to assume RoleB. ## Why it matters here Chains can lead to roles broader than the starting one. Review who can assume what, not only what each role allows. ## Related pages - [Privileged role: meaning and an example](/glossary/privileged-role/) - [Policy generation: meaning and an example](/glossary/policy-generation/) - [Job function policy: meaning and an example](/glossary/job-function-policy/) - [All of them](/glossary/) ## Sources - https://docs.aws.amazon.com/IAM/latest/UserGuide/id_roles_terms-and-concepts.html (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/glossary/role-chaining/ --- # Privileged role Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. A privileged role is a built-in Azure role whose permissions include broad management or the ability to assign access to others. Microsoft groups these roles in a Privileged category. Azure also lets an administrator use the elevateAccess action to gain User Access Administrator at tenant scope. ## An example, sourced Microsoft's privileged roles list shows Owner with the action "*", described as full access including the ability to assign roles, and Contributor, which grants full access to manage all resources but not to assign roles. ## Why it matters here Roles that can assign roles can widen their own access. Keep them rare and time-limited. ## Related pages - [Policy generation: meaning and an example](/glossary/policy-generation/) - [Job function policy: meaning and an example](/glossary/job-function-policy/) - [Policy inheritance: meaning and an example](/glossary/policy-inheritance/) - [All of them](/glossary/) ## Sources - https://learn.microsoft.com/en-us/azure/role-based-access-control/built-in-roles/privileged (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/glossary/privileged-role/ --- # Policy generation Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. Policy generation is the creation of an IAM policy from observed activity. IAM Access Analyzer reads AWS CloudTrail logs for a date range and builds a template of the permissions the entity actually used. AWS says the generated template covers actions and services found in the CloudTrail events it analyzed. ## An example, sourced AWS documentation says Access Analyzer reviews your CloudTrail logs and generates a policy template that contains the permissions the entity used in your specified date range. ## Why it matters here It supplies a starting point for a smaller policy. The result still needs review, since it reflects only the activity in the window. ## Related pages - [Job function policy: meaning and an example](/glossary/job-function-policy/) - [Policy inheritance: meaning and an example](/glossary/policy-inheritance/) - [Role assignment condition: meaning and an example](/glossary/role-assignment-condition/) - [All of them](/glossary/) ## Sources - https://docs.aws.amazon.com/IAM/latest/UserGuide/access-analyzer-policy-generation.html (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/glossary/policy-generation/ --- # Job function policy Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. A job function policy is an AWS managed policy that grants the permissions typical for a role in an organization, such as database administrator or security auditor. It is a starting point, not a minimal grant. AWS says they can be attached to any IAM identity. ## An example, sourced AWS documentation lists job function policies including Database administrator, Network administrator and Security auditor, and says that to move to least privilege you can run IAM Access Analyzer to monitor principals that use them. ## Why it matters here They are easier to review than custom policies, but they cover whole job functions. Use them to start and tighten with usage data. ## Related pages - [Policy inheritance: meaning and an example](/glossary/policy-inheritance/) - [Role assignment condition: meaning and an example](/glossary/role-assignment-condition/) - [Zone of trust: meaning and an example](/glossary/zone-of-trust/) - [All of them](/glossary/) ## Sources - https://docs.aws.amazon.com/IAM/latest/UserGuide/access_policies_job-functions.html (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/glossary/job-function-policy/ --- # Policy inheritance Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. Policy inheritance is the Google Cloud rule that a resource receives the allow policies of its parent. A grant at the organization or folder therefore reaches every project and resource beneath it. The union rule means a lower-level policy cannot take away what a parent granted. ## An example, sourced Google's documentation says resources inherit the allow policies of the parent resource, and that the effective allow policy for a resource is the union of the policy set at that resource and the one inherited from its parent. ## Why it matters here Grant at the lowest level that works. A role bound at the folder cannot be removed by editing a project. ## Related pages - [Role assignment condition: meaning and an example](/glossary/role-assignment-condition/) - [Zone of trust: meaning and an example](/glossary/zone-of-trust/) - [Long-term credentials: meaning and an example](/glossary/long-term-credentials/) - [All of them](/glossary/) ## Sources - https://cloud.google.com/iam/docs/resource-hierarchy-access-control (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/glossary/policy-inheritance/ --- # Role assignment condition Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. A role assignment condition is an extra check added to an Azure role assignment that tests attributes of the request, giving attribute-based access control on top of RBAC. It narrows what a granted role can do. The condition is stored with the assignment and checked when the action runs. ## An example, sourced Microsoft's documentation says Azure ABAC builds on Azure RBAC by adding role assignment conditions based on attributes in the context of specific actions, as an additional check on the assignment. ## Why it matters here A condition can narrow a data role to blobs with a matching tag without creating new roles. It reduces role sprawl. ## Related pages - [Zone of trust: meaning and an example](/glossary/zone-of-trust/) - [Long-term credentials: meaning and an example](/glossary/long-term-credentials/) - [Data action: meaning and an example](/glossary/data-action/) - [All of them](/glossary/) ## Sources - https://learn.microsoft.com/en-us/azure/role-based-access-control/conditions-overview (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/glossary/role-assignment-condition/ --- # Zone of trust Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. A zone of trust is the account or organization that an IAM Access Analyzer external access analyzer treats as safe. Access from principals inside it produces no finding, and access from outside it does. An external access analyzer is created for one organization or account at a time. ## An example, sourced AWS documentation says the resources and principals within the chosen organization or account comprise the zone of trust, and that sharing within it is considered safe so no finding is generated. ## Why it matters here Choosing an account as the zone instead of the organization changes which sharing is flagged, so the choice shapes the results. ## Related pages - [Long-term credentials: meaning and an example](/glossary/long-term-credentials/) - [Data action: meaning and an example](/glossary/data-action/) - [Assignable scope: meaning and an example](/glossary/assignable-scope/) - [All of them](/glossary/) ## Sources - https://docs.aws.amazon.com/IAM/latest/UserGuide/access-analyzer-findings.html (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/glossary/zone-of-trust/ --- # Long-term credentials Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. Long-term credentials are keys that stay valid until someone deletes or deactivates them, such as AWS IAM user access keys. They are the opposite of temporary credentials. AWS says IAM users with access keys are an account security risk that needs careful handling. ## An example, sourced AWS documentation describes an access key as an access key ID such as AKIAIOSFODNN7EXAMPLE plus a secret access key, says the secret can be retrieved only at creation, and says an IAM user can have a maximum of two. ## Why it matters here Every long-term key is standing access that can leak. Roles with temporary credentials remove the need for most of them. ## Related pages - [Data action: meaning and an example](/glossary/data-action/) - [Assignable scope: meaning and an example](/glossary/assignable-scope/) - [Access debt: meaning and an example](/glossary/access-debt/) - [All of them](/glossary/) ## Sources - https://docs.aws.amazon.com/IAM/latest/UserGuide/id_credentials_access-keys.html (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/glossary/long-term-credentials/ --- # Data action Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. A data action is an Azure permission for operations on the data inside a resource, such as reading a blob, as opposed to managing the resource itself. Role definitions list them under dataActions. Reading a blob is a data action, while creating a storage account is a management action. ## An example, sourced Microsoft's documentation says DataActions is an array of strings that specifies the data plane actions the role allows on data within an object, and the Storage Blob Data Reader role reads and lists containers and blobs. ## Why it matters here Management and data access are separate grants. A role that manages a storage account does not need to read its contents. ## Related pages - [Assignable scope: meaning and an example](/glossary/assignable-scope/) - [Access debt: meaning and an example](/glossary/access-debt/) - [Right-sizing: meaning and an example](/glossary/right-sizing/) - [All of them](/glossary/) ## Sources - https://learn.microsoft.com/en-us/azure/role-based-access-control/role-definitions (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/glossary/data-action/ --- # Assignable scope Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. An assignable scope is a field in an Azure role definition that lists where the role can be assigned, such as subscriptions or management groups. A custom role cannot be assigned outside those scopes. It also documents where the owner of the role expects it to be used. ## An example, sourced Microsoft's custom role example lists assignable scopes of two subscriptions and one management group, in the form /subscriptions/{subscriptionId1}. ## Why it matters here It limits where a custom role can spread. A role limited to one subscription cannot be assigned in others. ## Related pages - [Access debt: meaning and an example](/glossary/access-debt/) - [Right-sizing: meaning and an example](/glossary/right-sizing/) - [Role-based access control: meaning and an example](/glossary/role-based-access-control/) - [All of them](/glossary/) ## Sources - https://learn.microsoft.com/en-us/azure/role-based-access-control/custom-roles (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/glossary/assignable-scope/ --- # What changed in the last 90 days Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. Dated notes on changes that affect least privilege in the cloud: new provider features, changes to documentation, and updates to NIST controls. Each item names the source page, says what changed and when it was read, and links to the guide that the change affects. - [AWS introduces account access manager for assigning IAM roles to workforce users](/news/aws-account-access-manager/): 5 Nov 2026 - [Google Cloud basic roles Admin, Writer and Reader are generally available](/news/google-basic-roles-admin-writer-reader-ga/): 29 Oct 2026 - [Google Cloud can revoke Workforce Identity Federation sessions in general availability](/news/google-workforce-session-revocation-ga/): 6 Oct 2026 - [IAM Identity Center adds network access controls for Identity Store](/news/identity-center-identity-store-network-controls/): 5 Oct 2026 - [Policy Troubleshooter now explains access for agent identities](/news/google-policy-troubleshooter-agent-identities/): 15 Sep 2026 - [Amazon ECS extends CPU and memory condition keys to RunTask and StartTask](/news/ecs-cpu-memory-condition-keys-runtask/): 10 Sep 2026 - [Google Cloud IAM MCP server reaches general availability for custom roles and deny policies](/news/google-iam-mcp-server-ga/): 10 Sep 2026 - [Google Cloud Policy Assist API returns IAM role suggestions in Preview](/news/google-policy-assist-api-preview/): 9 Sep 2026 - [Google Cloud Policy Analyzer MCP server is generally available](/news/google-policy-analyzer-mcp-server-ga/): 1 Sep 2026 - [Lambda functions support full IAM resource-based policies](/news/lambda-full-iam-resource-based-policies/): 25 Aug 2026 - [IAM raises the default managed policies per role from 10 to 20](/news/iam-20-managed-policies-per-role/): 19 Aug 2026 - [IAM Policy Autopilot generates policies from Terraform plan files](/news/iam-policy-autopilot-terraform-plan/): 18 Aug 2026 ## Related pages - [Least privilege for 20 cloud services](/services/) - [Routes to least privilege: tools and platforms](/routes/) - [The least privilege data note: every edition](/data-note/) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/news/ --- # AWS introduces account access manager for assigning IAM roles to workforce users Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. The AWS Security Blog described account access manager on 5 November 2026. AWS says it assigns existing IAM roles from individual accounts to users and groups in IAM Identity Center, keeps role tags, paths and trust policy conditions, and records the individual user in each role session for audit. The AWS Security Blog post of 5 November 2026 says account access manager lets administrators create an entitlement that links an IAM Identity Center user or group to a specific IAM role in an account. Users reach the role from a portal or with aws login on the command line. AWS says each role session carries user context through trusted identity propagation, so actions are attributed to the individual even when several users assume the same role. It lists attribute-based access control, trust policy conditions on network or identity provider claims, and role paths as capabilities kept. It exposes its own account-access API namespace with separate permissions and CloudTrail events. ## Related pages - [Google Cloud basic roles Admin, Writer and Reader are generally available](/news/google-basic-roles-admin-writer-reader-ga/) - [Google Cloud can revoke Workforce Identity Federation sessions in general availability](/news/google-workforce-session-revocation-ga/) - [IAM Identity Center adds network access controls for Identity Store](/news/identity-center-identity-store-network-controls/) - [All of them](/news/) ## Sources - https://aws.amazon.com/blogs/security/introducing-aws-account-access-manager-iam-role-flexibility-with-centralized-workforce-access/ (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/news/aws-account-access-manager/ --- # Google Cloud basic roles Admin, Writer and Reader are generally available Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. Google Cloud IAM release notes dated 29 October 2026 say the Admin, Writer and Reader basic roles are generally available. Google recommends them over the legacy Owner, Editor and Viewer roles when a basic role must be granted, because the newer roles support IAM Conditions. The IAM release notes of 29 October 2026 say the basic roles Admin (roles/admin), Writer (roles/writer) and Reader (roles/reader) are generally available. The legacy Owner, Editor and Viewer roles remain available and are being actively maintained. Google says that if you need to grant a basic role, it recommends granting Admin, Writer or Reader instead of a legacy basic role, because the newer roles support additional features such as IAM Conditions. Basic roles remain broad, so a condition is the one way to narrow one when no predefined role fits. ## Related pages - [Google Cloud can revoke Workforce Identity Federation sessions in general availability](/news/google-workforce-session-revocation-ga/) - [IAM Identity Center adds network access controls for Identity Store](/news/identity-center-identity-store-network-controls/) - [Policy Troubleshooter now explains access for agent identities](/news/google-policy-troubleshooter-agent-identities/) - [All of them](/news/) ## Sources - https://cloud.google.com/iam/docs/release-notes (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/news/google-basic-roles-admin-writer-reader-ga/ --- # Google Cloud can revoke Workforce Identity Federation sessions in general availability Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. Google Cloud IAM release notes dated 6 October 2026 say revoking active sessions and short-lived credentials for Workforce Identity Federation users is generally available across all clients. Google says administrators can revoke workforce user sessions with the gcloud CLI or the REST API. The IAM release notes of 6 October 2026 say revoking active sessions and short-lived credentials for Workforce Identity Federation principals across all clients is generally available. Google says sessions can be revoked with the gcloud CLI or the REST API. Short-lived credentials limit how long a stolen token is useful, and revocation shortens that further after a user is removed or a role is cut. Google points to the Revoke Workforce Identity Federation user sessions page for the steps. ## Related pages - [IAM Identity Center adds network access controls for Identity Store](/news/identity-center-identity-store-network-controls/) - [Policy Troubleshooter now explains access for agent identities](/news/google-policy-troubleshooter-agent-identities/) - [Amazon ECS extends CPU and memory condition keys to RunTask and StartTask](/news/ecs-cpu-memory-condition-keys-runtask/) - [All of them](/news/) ## Sources - https://cloud.google.com/iam/docs/release-notes (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/news/google-workforce-session-revocation-ga/ --- # IAM Identity Center adds network access controls for Identity Store Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. AWS announced on 5 October 2026 that IAM Identity Center supports network access controls for Identity Store. AWS says administrators can limit the Identity Store API and the SCIM API to allowed VPC endpoints, source VPCs or IP ranges, with different rules per API. The controls are optional and off by default. AWS What's New, posted 5 October 2026, says the Identity Store API can be required to arrive through allowed VPC endpoints in an account or organization, or from specific source VPCs. For both the Identity Store API and the SCIM API, requests can be allowed only from specific IP ranges. AWS says the controls are optional and turned off by default, requests that AWS services make on your behalf are exempt, and configuration is done through the Identity Store API using the AWS SDKs and CLI. The capability is available in all Regions where IAM Identity Center is offered. ## Related pages - [Policy Troubleshooter now explains access for agent identities](/news/google-policy-troubleshooter-agent-identities/) - [Amazon ECS extends CPU and memory condition keys to RunTask and StartTask](/news/ecs-cpu-memory-condition-keys-runtask/) - [Google Cloud IAM MCP server reaches general availability for custom roles and deny policies](/news/google-iam-mcp-server-ga/) - [All of them](/news/) ## Sources - https://aws.amazon.com/about-aws/whats-new/2026/10/aws-identity-store-network-controls/ (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/news/identity-center-identity-store-network-controls/ --- # Policy Troubleshooter now explains access for agent identities Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. Google Cloud Policy Intelligence release notes dated 15 September 2026 say Policy Troubleshooter supports agent identities. Google says it can troubleshoot IAM allow policies, deny policies and principal access boundary policies from an agent principal identifier or from an error ID in an access denial event. The Policy Intelligence release notes of 15 September 2026 say Policy Troubleshooter now supports troubleshooting access for agent identities acting under their own authority. A user enters the agent principal identifier, or uses an error ID from an access denial event. Google lists three policy types the troubleshooter evaluates: IAM allow policies, deny policies and principal access boundary policies. Seeing which of the three blocked or allowed a request is the practical step in tightening a role without breaking a workload. ## Related pages - [Amazon ECS extends CPU and memory condition keys to RunTask and StartTask](/news/ecs-cpu-memory-condition-keys-runtask/) - [Google Cloud IAM MCP server reaches general availability for custom roles and deny policies](/news/google-iam-mcp-server-ga/) - [Google Cloud Policy Assist API returns IAM role suggestions in Preview](/news/google-policy-assist-api-preview/) - [All of them](/news/) ## Sources - https://cloud.google.com/policy-intelligence/docs/release-notes (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/news/google-policy-troubleshooter-agent-identities/ --- # Amazon ECS extends CPU and memory condition keys to RunTask and StartTask Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. AWS announced on 10 September 2026 that the ecs:task-cpu and ecs:task-memory IAM condition keys now apply to the RunTask and StartTask APIs. AWS says administrators can enforce the same CPU and memory limits however a task is launched, at no additional cost in all Regions where ECS is available. AWS What's New, posted 10 September 2026, says the ecs:task-cpu and ecs:task-memory condition keys were available only on RegisterTaskDefinition, CreateService and UpdateService. They are now also evaluated on RunTask and StartTask. AWS frames the change as a single mechanism to control resource allocation across ECS environments and to prevent unexpected cost overruns. For access control, it means a policy that caps task size no longer has a gap for tasks started directly, and the change is available in all Regions where ECS is available. ## Related pages - [Google Cloud IAM MCP server reaches general availability for custom roles and deny policies](/news/google-iam-mcp-server-ga/) - [Google Cloud Policy Assist API returns IAM role suggestions in Preview](/news/google-policy-assist-api-preview/) - [Google Cloud Policy Analyzer MCP server is generally available](/news/google-policy-analyzer-mcp-server-ga/) - [All of them](/news/) ## Sources - https://aws.amazon.com/about-aws/whats-new/2026/09/amazon-ecs-expands-condition-key-support/ (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/news/ecs-cpu-memory-condition-keys-runtask/ --- # Google Cloud IAM MCP server reaches general availability for custom roles and deny policies Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. Google Cloud IAM release notes dated 10 September 2026 say the IAM remote MCP server is generally available. Google says AI applications can connect to it to inspect and manage custom roles and deny policies across resources, two of the main tools for narrowing access below a predefined role. The IAM release notes of 10 September 2026 say the Identity and Access Management Model Context Protocol server is generally available. Google says you can connect to the IAM remote MCP server from AI applications to inspect and manage custom roles and deny policies across your resources. Custom roles let an administrator list only the permissions a task needs, and deny policies block named permissions regardless of the roles granted. The release note describes a new way to reach both through a tool interface and links to the usage and reference pages. ## Related pages - [Google Cloud Policy Assist API returns IAM role suggestions in Preview](/news/google-policy-assist-api-preview/) - [Google Cloud Policy Analyzer MCP server is generally available](/news/google-policy-analyzer-mcp-server-ga/) - [Lambda functions support full IAM resource-based policies](/news/lambda-full-iam-resource-based-policies/) - [All of them](/news/) ## Sources - https://cloud.google.com/iam/docs/release-notes (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/news/google-iam-mcp-server-ga/ --- # Google Cloud Policy Assist API returns IAM role suggestions in Preview Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. Google Cloud release notes dated 9 September 2026 say the Policy Assist REST API and a Policy Assist remote MCP server are in Preview. Google says Policy Assist returns IAM role suggestions for individual principals with AI assistance, so teams can request a narrower predefined role by code instead of in the console. The Policy Intelligence release notes of 9 September 2026 list the Policy Assist REST API and the Policy Assist remote MCP server as Preview features. Google says Policy Assist lets you get IAM role suggestions for individual principals with AI assistance. The IAM release notes of the same date say role suggestions from Gemini can be obtained programmatically through the Policy Assist API, in Preview. Preview features carry the terms Google attaches to Preview launches, so a production workflow should treat the suggestions as input to review, not as an approved role. ## Related pages - [Google Cloud Policy Analyzer MCP server is generally available](/news/google-policy-analyzer-mcp-server-ga/) - [Lambda functions support full IAM resource-based policies](/news/lambda-full-iam-resource-based-policies/) - [IAM raises the default managed policies per role from 10 to 20](/news/iam-20-managed-policies-per-role/) - [All of them](/news/) ## Sources - https://cloud.google.com/policy-intelligence/docs/release-notes (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/news/google-policy-assist-api-preview/ --- # Google Cloud Policy Analyzer MCP server is generally available Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. Google Cloud release notes dated 1 September 2026 say the Policy Analyzer MCP server is generally available. Google says it lets agents and AI applications analyze and audit IAM configurations, which extends Policy Analyzer, the tool for finding which principals can access which resources, to programmatic callers. The Policy Intelligence release notes, dated 1 September 2026, say the Policy Analyzer MCP server is generally available. Google says it lets agents and AI applications analyze and audit IAM configurations. Policy Analyzer is the Policy Intelligence tool for checking which principals have access to which resources. The release note points to the documentation page for using the Policy Analyzer remote MCP server and does not describe new analysis types, so the change is about how the existing analysis can be called. ## Related pages - [Lambda functions support full IAM resource-based policies](/news/lambda-full-iam-resource-based-policies/) - [IAM raises the default managed policies per role from 10 to 20](/news/iam-20-managed-policies-per-role/) - [IAM Policy Autopilot generates policies from Terraform plan files](/news/iam-policy-autopilot-terraform-plan/) - [All of them](/news/) ## Sources - https://cloud.google.com/policy-intelligence/docs/release-notes (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/news/google-policy-analyzer-mcp-server-ga/ --- # Lambda functions support full IAM resource-based policies Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. AWS announced on 25 August 2026 that Lambda functions now support full IAM resource-based policies. AWS says one policy document can cover multiple principals and actions and use the full range of IAM condition keys, such as source IP or principal tag, at no additional charge in all commercial Regions. AWS What's New, posted 25 August 2026, says Lambda previously required permissions to be added one principal at a time. Now a single resource-based policy can define permissions for several principals and actions and use the full range of IAM condition keys. AWS gives examples of restricting access by source IP or principal tag. Policies can be updated in one step in the Lambda console JSON editor, the AWS CLI, an SDK, or infrastructure as code such as CloudFormation and AWS SAM. AWS says the capability is available in all commercial Regions at no additional charge. ## Related pages - [IAM raises the default managed policies per role from 10 to 20](/news/iam-20-managed-policies-per-role/) - [IAM Policy Autopilot generates policies from Terraform plan files](/news/iam-policy-autopilot-terraform-plan/) - [Least privilege data note, November 2026](/data-note/2026-11/) - [All of them](/news/) ## Sources - https://aws.amazon.com/about-aws/whats-new/2026/08/aws-lambda-full-iam-resource-based-policies/ (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/news/lambda-full-iam-resource-based-policies/ --- # IAM raises the default managed policies per role from 10 to 20 Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. AWS announced on 19 August 2026 that the default quota for managed policies attached to one IAM role rose from 10 to 20. AWS says teams can now split permissions into purpose-specific policies without a quota request, and can request up to 25 through Service Quotas if they need more. AWS What's New, posted 19 August 2026, says the default quota for managed policies per role is now 20, up from 10. The change applies automatically to all IAM roles in an account in commercial Regions, AWS GovCloud (US) and China Regions, with no action required. AWS gives the reason as fewer Service Quota requests when following best practices such as separating permissions into purpose-specific policies. For least privilege work, smaller policies with one job each are easier to review and remove than one large policy, and the higher default removes one reason to merge them. ## Related pages - [IAM Policy Autopilot generates policies from Terraform plan files](/news/iam-policy-autopilot-terraform-plan/) - [Least privilege data note, November 2026](/data-note/2026-11/) - [Least privilege data note, October 2026](/data-note/2026-10/) - [All of them](/news/) ## Sources - https://aws.amazon.com/about-aws/whats-new/2026/08/aws-iam-quota-increase/ (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/news/iam-20-managed-policies-per-role/ --- # IAM Policy Autopilot generates policies from Terraform plan files Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. AWS announced on 18 August 2026 that IAM Policy Autopilot, an open source tool, can now generate baseline IAM policies from a Terraform plan file. AWS says the policies cover the create, read, update and delete calls for the resources in the plan and name resource ARNs instead of wildcards where possible. AWS What's New, posted 18 August 2026, says IAM Policy Autopilot can pass a Terraform plan file to a deterministic analysis that produces a policy scoped to the CRUD functions of the resources in that plan. Until now the tool analyzed application source code only, which left the permissions needed to deploy infrastructure unaddressed. AWS says the generated policies reference specific resource ARNs rather than wildcards, when possible, and that the tool runs on your own machine at no additional cost. AWS describes the output as a baseline to refine as the application evolves, so the policy still needs review against what the deployment role actually calls. ## Related pages - [Least privilege data note, November 2026](/data-note/2026-11/) - [Least privilege data note, October 2026](/data-note/2026-10/) - [Least privilege data note, September 2026](/data-note/2026-09/) - [All of them](/news/) ## Sources - https://aws.amazon.com/about-aws/whats-new/2026/08/iam-policy-autopilot-now-supports-terraform-plan-files (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/news/iam-policy-autopilot-terraform-plan/ --- # 30 questions, answered from the data Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. Answers to common questions about least privilege in the cloud: what the principle means, how it differs from zero trust and RBAC, how to find unused access, and how to start. Each answer is short, names the provider documentation behind it and gives the date it was read. **What is the principle of least privilege in cloud?** In cloud, it means each identity gets only the actions on the resources and conditions a task needs. NIST's glossary defines [least privilege](https://csrc.nist.gov/glossary/term/least_privilege) as restricting the access privileges of users, or processes acting for them, to the minimum necessary to accomplish assigned tasks. The control that requires it is NIST SP 800-53 [AC-6](https://csrc.nist.gov/projects/cprt/catalog#/cprt/framework/version/SP_800_53_5_1_1/home?element=AC-6). See [the principle](/principle-of-least-privilege/) for the full reading. **What is least privilege access?** Least privilege access is access limited to the minimum needed for assigned work. NIST's glossary uses that wording and applies it to users and to processes acting on behalf of users, so it covers roles, keys and services, not only people. AWS documentation says to define the actions on specific resources under specific conditions, and calls the result [least-privilege permissions](https://docs.aws.amazon.com/IAM/latest/UserGuide/best-practices.html). **Is zero trust the same as least privilege?** No. NIST's glossary defines [zero trust](https://csrc.nist.gov/glossary/term/zero_trust) as a set of concepts for enforcing accurate, least privilege per-request access decisions when the network is viewed as compromised, drawing on NIST SP 800-207. Least privilege is therefore one property that zero trust asks for on every request. A system can restrict privileges and still trust the network, so the two are related but not identical. **What is the difference between least privilege and need to know?** Need to know is a decision that a recipient requires access to specific information to carry out official duties. NIST's glossary defines [need to know](https://csrc.nist.gov/glossary/term/need_to_know) that way. Least privilege limits the system privileges a user or process holds to the minimum for assigned tasks. In practice, need to know decides which information, and least privilege decides which actions and resources. **What is the difference between least privilege and separation of duties?** Least privilege limits how much one identity can do; separation of duties limits what one identity can do alone. NIST's [separation of duty](https://csrc.nist.gov/glossary/term/separation_of_duty) entry says no user should have enough privileges to misuse the system on their own, and gives the example of the person authorizing a paycheck not also preparing it. **Is least privilege the same as role-based access control?** No. Role-based access control is a way to assign permissions by job function, as the NIST glossary entry for [RBAC](https://csrc.nist.gov/glossary/term/role_based_access_control) describes. Least privilege is the goal that each role carries only what its function needs. A role can follow RBAC and still hold far more than the job requires, which is why the permissions inside each role need review. **What is the difference between RBAC and ABAC for least privilege?** RBAC grants permissions by job function, while ABAC grants them by attributes. AWS documentation says AWS calls these attributes tags, and that the traditional IAM model is [RBAC compared with ABAC](https://docs.aws.amazon.com/IAM/latest/UserGuide/introduction_attribute-based-access-control.html). ABAC can reduce the number of roles to maintain, but tag permissions need the same least privilege care as any other permission. **What is privilege creep?** Privilege creep is the slow accumulation of access beyond what a person or workload needs, usually because new access is added for new tasks and old access is never removed. It is the reason least privilege needs repeat checks, not a one-time clean-up. See [privilege creep](/privilege-creep/) for the causes and the checks that catch it. **How do I implement least privilege in AWS?** Start broad, then narrow with evidence. AWS documentation says you might begin with broad permissions while you learn what a workload needs, then reduce them as the use case matures, and that [AWS managed policies](https://docs.aws.amazon.com/IAM/latest/UserGuide/best-practices.html) may not be least privilege for your case. Replace them with customer managed policies, generate policies from activity, and remove unused access. **How do I implement least privilege in Azure?** Grant roles at the narrowest scope. Microsoft's Azure RBAC documentation says to grant users the least privilege to get their work done, to avoid broader roles at broader scopes, and to use a narrow scope such as a resource group or resource for any [privileged administrator role](https://learn.microsoft.com/en-us/azure/role-based-access-control/best-practices). Custom roles should include only the permissions users need. **How do I implement least privilege in Google Cloud?** Prefer predefined or custom roles over basic roles, and grant them low in the resource hierarchy. Google's documentation says [basic roles](https://cloud.google.com/iam/docs/roles-overview) are highly permissive and give broad access to resources. Then use the IAM recommender, which compares granted permissions with what a principal used in the last 90 days, to remove the rest. **How do I find unused permissions in AWS?** Use IAM Access Analyzer's unused access findings and last accessed information. AWS documentation says an unused access analyzer reports [unused roles, unused IAM user access keys and passwords, and unused permissions](https://docs.aws.amazon.com/IAM/latest/UserGuide/access-analyzer-findings.html) in your organization and accounts. Last accessed information shows when an identity last used an AWS service or action, so you can remove what was never touched. **Can AWS generate a least privilege policy for me?** Yes, from past activity. AWS documentation says IAM Access Analyzer reviews your CloudTrail logs and generates a policy containing the permissions an entity used in your specified date range, up to 90 days. You can replace a broad policy with the generated one after testing it. See [policy generation](https://docs.aws.amazon.com/IAM/latest/UserGuide/access-analyzer-policy-generation.html) for the steps. **What does the IAM recommender do in Google Cloud?** It suggests removing or replacing roles that grant permissions a principal does not use. Google's documentation says the recommender builds policy insights by comparing a principal's permissions with those used in the last 90 days. A person reviews each recommendation and applies or dismisses it; it is described in [role recommendations](https://cloud.google.com/iam/docs/recommender-overview). **Why should I avoid wildcards in IAM policies?** A wildcard grants more actions or resources than a task needs, and new actions can match it later. AWS documentation says to define actions on specific resources, and the [policy summary access levels](https://docs.aws.amazon.com/IAM/latest/UserGuide/access_policies_understand-policy-summary-access-level-summaries.html) page lets you see whether a policy grants List, Read, Write, Permissions management or Tagging actions, which shows how wide a wildcard reaches. **Why are basic roles like Owner and Editor risky?** They grant broad access across most services. Google's documentation says [basic roles](https://cloud.google.com/iam/docs/roles-overview) are highly permissive. Google's service account guidance also notes that some default service accounts were automatically granted the Editor role on a project, which allows reading and modifying most resources. Use a predefined or custom role for the task instead. **What should I remove first when reducing access?** Remove what shows no activity and what can change access for others. AWS documentation says to [regularly review and remove unused](https://docs.aws.amazon.com/IAM/latest/UserGuide/best-practices.html) users, roles, permissions, policies and credentials. Microsoft's access reviews documentation separately advises checking how many users hold administrative roles. Unused credentials and unused privileged roles are the usual first targets. **How often should I review access?** Set a repeating schedule instead of a single review. Microsoft's access reviews documentation says you can have reviews recur weekly, monthly, quarterly or annually, with reviewers notified at the start of each. AWS tells you to [regularly review](https://docs.aws.amazon.com/IAM/latest/UserGuide/best-practices.html) and remove unused access. Privileged roles suit the shortest interval. **What is an access review?** An access review is a periodic check by a named reviewer on whether each person still needs the access they hold. Microsoft's [access reviews overview](https://learn.microsoft.com/en-us/entra/id-governance/access-reviews-overview) describes setting up recurring reviews where reviewers approve or deny access, and says to check how many users hold administrative roles and whether guests were removed. **Should I use temporary credentials instead of access keys?** Yes where you can. AWS documentation says to use temporary security credentials, such as IAM roles, instead of long-term [access keys](https://docs.aws.amazon.com/IAM/latest/UserGuide/id_credentials_access-keys.html), which belong to an IAM user or the root user. Google likewise says service account keys can become a security risk and recommends alternatives. Temporary credentials expire by design. **How should I rotate access keys?** Avoid long-term keys first; rotate the ones you must keep. AWS documentation says to update access keys when needed for use cases that require long-term credentials, and an IAM user can have a maximum of two access keys, which makes rotation possible without downtime. Google's guidance on [service account keys](https://cloud.google.com/iam/docs/best-practices-for-managing-service-account-keys) lists rotation to reduce the risk of leaked keys. **How do I require MFA in an IAM policy?** Add the aws:MultiFactorAuthPresent condition key. AWS documentation shows a Bool condition with "aws:MultiFactorAuthPresent": "true" to allow actions only when the request was authenticated with MFA, in the page on [requiring MFA for API access](https://docs.aws.amazon.com/IAM/latest/UserGuide/id_credentials_mfa_configure-api-require.html). AWS also recommends phishing-resistant MFA such as passkeys and security keys. **What is just in time access for least privilege?** It limits when a privilege exists, not only what it is. Microsoft's PIM documentation says an eligible assignment must be activated, for example with an MFA check or approval, and that [time-bound assignments](https://learn.microsoft.com/en-us/entra/id-governance/privileged-identity-management/pim-configure) use start and end dates. The access is the same while active; it is absent the rest of the time. **How do service accounts fit with least privilege?** Give each workload its own service account with only the roles it needs, and avoid keys. Google's [service account best practices](https://cloud.google.com/iam/docs/best-practices-service-accounts) warn that default service accounts were sometimes granted the Editor role automatically. Microsoft says managed identities remove the need to handle credentials. See the [Google Cloud service identities guide](/guides/google-cloud-service-identities/). **What is workload identity federation?** It lets a workload outside the cloud get short-lived access without a stored key. Google's [workload identity federation](https://cloud.google.com/iam/docs/workload-identity-federation) documentation describes exchanging an external identity for Google Cloud access, and Microsoft's documentation describes the same pattern for Entra. It supports least privilege by removing long-lived secrets. See the [Azure workload identities guide](/guides/azure-workload-identities/). **How do I limit third-party access to my cloud account?** Give the third party a role that only they can assume, with narrow permissions and conditions. AWS documentation on [third-party access](https://docs.aws.amazon.com/IAM/latest/UserGuide/id_roles_common-scenarios_third-party.html) describes granting a third party access through a role in your account. Review the permissions on each role. See the [AWS third-party identities guide](/guides/aws-third-party-identities/). **What is a data perimeter in AWS?** It is a set of always-on guardrails around identities, resources and networks. AWS's whitepaper [Building a Data Perimeter on AWS](https://docs.aws.amazon.com/whitepapers/latest/building-a-data-perimeter-on-aws/building-a-data-perimeter-on-aws.html), published 13 June 2023, says perimeter controls help protect against unintended access and configuration errors. It limits who and what can reach data even when a policy is too broad. **What are examples of least privilege in AWS, Azure and Google Cloud?** Each provider documents narrow patterns per service, and the service pages here list the exact action and role names for 20 services across AWS, Azure and Google Cloud. Start with [Amazon S3](/services/aws-s3/), [Azure Key Vault](/services/azure-key-vault/) and [Cloud Storage](/services/gcp-cloud-storage/). Each page gives a minimal policy or role binding for a common task. **Do conditions make a broad role safer?** They narrow it, and the role stays broad outside the condition. Azure's documentation says role assignment conditions add a check to a role assignment for finer-grained access control, and Google's documentation describes [IAM Conditions](https://cloud.google.com/iam/docs/conditions-overview) that grant a role only when an expression is true. Use them with the narrowest role that fits. **What is a deny policy and when does it help?** A deny policy blocks named permissions for named principals no matter which roles they hold. Google's [deny policies](https://cloud.google.com/iam/docs/deny-overview) page describes setting guardrails that override allow policies. Deny rules suit limits that must hold across many roles, such as stopping a group from ever deleting logs, while allow policies grant the specific access. ## Related pages - [The Access Debt Calculator: the data, downloadable](/datasets/) - [About](/about/) - [Editorial policy](/editorial-policy/) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/faq/ --- # The Access Debt Calculator: how big your least privilege clean-up is Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. The Access Debt Calculator turns counts from your estate (identities, identities with unused permissions, permissions granted and used per identity, accounts, changes shipped per week and risk classes) into the size of two plans: a permission-by-permission clean-up and a boundary plan. It states each assumption and shows the arithmetic. Every file below is this site's own data, edition 16 Nov 2026, under CC BY 4.0. The [methodology](/methodology/) explains how it was made. ## Downloads - [Access Debt Calculator assumptions (CSV)](/datasets/access-debt-calculator-assumptions-2026-11-16.csv): 8 rows, edition 16 Nov 2026 - [Access Debt Calculator assumptions (JSON)](/datasets/access-debt-calculator-assumptions-2026-11-16.json): Every rule the calculator applies, with the default example worked through. - [Least privilege routes directory (CSV)](/datasets/least-privilege-routes-2026-11-16.csv): 24 rows, edition 16 Nov 2026 - [Least privilege routes directory (JSON)](/datasets/least-privilege-routes-2026-11-16.json): Provider tools, right-sizing products and boundary platforms, the same fields for each, alphabetical, with source and read date. ## Cite this **Cite this.** Cite this line: 'Least Privilege Principle, The Access Debt Calculator, edition 16 Nov 2026, https://leastprivilegeprinciple.com/methodology/' ## Changelog - 16 Nov 2026: first edition. Every row read from the source linked in the row on 16 Nov 2026. ## Related pages - [About](/about/) - [Editorial policy](/editorial-policy/) - [Corrections](/corrections/) ## Sources - AWS Organizations: service control policies: https://docs.aws.amazon.com/organizations/latest/userguide/orgs_manage_policies_scps.html (read 16 Nov 2026) - Act Security: https://act.security/platform (read 16 Nov 2026) - Apono: https://www.apono.io/ (read 16 Nov 2026) - AWS Control Tower: https://docs.aws.amazon.com/controltower/latest/userguide/pricing.html (read 10 Nov 2026) - AWS IAM Access Analyzer: https://aws.amazon.com/iam/access-analyzer/pricing/ (read 10 Nov 2026) - AWS IAM Identity Center temporary elevated access (TEAM): https://aws-samples.github.io/iam-identity-center-team/docs/overview/cost.html (read 16 Nov 2026) - Britive: https://www.britive.com/ (read 16 Nov 2026) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/datasets/ --- # A field guide to least privilege in AWS, Azure and Google Cloud, with the working shown Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. Least Privilege Principle is a field guide to least privilege in AWS, Azure and Google Cloud. It has a definition page built on NIST SP 800-53 control AC-6, guides to privilege creep, boundaries and a first 90 days, a policy template, and the Access Debt Calculator. Every fact carries a source and a read date. The publication has one subject: how an organization gets from broad cloud access to the access its work needs. The pages define the principle, show before and after permissions in each cloud, and explain where projects stop. The calculator sizes the clean-up from the counts a team can read from its own provider reports. The method is plain. Provider facts come from AWS, Microsoft and Google Cloud documentation, and standards text comes from NIST. Each page lists its sources with the date they were read, which is 16 Nov 2026 for the first edition. A statement a provider makes about itself is labeled as the provider's. When a feature was looked for and not found, the page says none found on that date. The site ranks no product and gives no score. Product reviews, workforce identity, cloud posture rankings and compliance crosswalks are outside its scope, and a page that touches them says so in one sentence. The calculator's inputs, assumptions and arithmetic are published on the methodology page so a reader can reproduce any result. ## Related pages - [Editorial policy](/editorial-policy/) - [Corrections](/corrections/) - [Contact](/contact/) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/about/ --- # How Least Privilege Principle is written Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. Least Privilege Principle publishes guides on cloud least privilege, each tied to provider documentation or a standard and dated. Every page answers its question in a short summary under the headline, names its numbers, and lists its sources with the date they were read, so a reader can check any claim in minutes. Facts come from primary documents: provider documentation, NIST publications and the text of the controls themselves. Where a provider describes its own product, the page says whose statement it is. Where a feature could not be found in the pages read, the page says none found, with the date, and does not claim the feature does not exist. Headlines state the answer in plain words. Numbers carry the unit, the source and the date. Examples are marked as illustrative when they are constructed, and sourced when they are quoted. Pages avoid promotion and avoid advice that the cited documents do not support. Each page shows when it was last reviewed. Pages are reviewed when a cited document changes, when a reader reports an error, and at least once every six months. A material change is recorded on the corrections page with its date. The glossary and the FAQ follow the same rules as the guides. Definitions stand alone in one sentence, and each is tied to a source page that was read and dated. ## Related pages - [Corrections](/corrections/) - [Contact](/contact/) - [Privacy](/privacy/) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/editorial-policy/ --- # Corrections and the correction rule Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. Corrections go to editor@leastprivilegeprinciple.com. Include the page address, the sentence you think is wrong and the document that shows the right figure. A confirmed error is fixed on the page, the review date changes, and a dated note describing the change is added to the corrections log. Corrections are checked against the cited source and against the current version of the provider or standards document. If a provider changes its documentation after a page was read, the page is updated and the new read date is shown. Readers can reproduce any calculator result from the published inputs and assumptions, and report arithmetic errors the same way. ## Related pages - [Contact](/contact/) - [Privacy](/privacy/) - [Terms](/terms/) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/corrections/ --- # Contact Least Privilege Principle Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. Write to editor@leastprivilegeprinciple.com with corrections, questions about a page, or requests to cite or reuse the Access Debt Calculator. Please include the page address and, for a correction, the sentence and the source that shows the right figure. Messages are read by the editor. The site has no accounts, no comment sections and no contact form. Email is the only channel, which keeps the record of every correction in one place. If a message concerns a security problem on the site itself, say so in the subject line. ## Related pages - [Privacy](/privacy/) - [Terms](/terms/) - [Access Debt Calculator methodology](/methodology/) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/contact/ --- # Privacy Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. Least Privilege Principle sets no cookies and runs no analytics script. The Access Debt Calculator runs in your browser and sends nothing: the numbers you enter stay on your device and are not stored. The web server keeps ordinary access logs, and nothing else is collected about visitors. The access logs hold the address, the page requested, the time and the browser string, and are used to operate the site and find faults. Email sent to the editor address is kept to answer it and to record corrections. There are no profiles, no advertising and no sale of data, so there is nothing to request or delete beyond an email you sent. ## Related pages - [Terms](/terms/) - [Access Debt Calculator methodology](/methodology/) - [30 questions about least privilege in the cloud](/faq/) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/privacy/ --- # Terms of use Facts read 16 Nov 2026. Page reviewed 16 Nov 2026. The content of Least Privilege Principle is provided for general information about cloud access control. It is not legal, audit or professional advice, and the pages describe provider documentation as read on the date shown. Check the current provider documentation before you change a production policy. You may quote short passages with a link to the page and its date. The Access Debt Calculator may be cited using the line on its page. The policy template is offered for adoption and adaptation by your organization, and you remain responsible for its fit to your environment and for testing changes before use. Provider and standards names belong to their owners and are used to identify the documents read. Quotations from NIST publications are public domain; quotations from provider documentation are short and attributed. The site is provided as is, and its pages may change as the documentation they cite changes. ## Related pages - [Access Debt Calculator methodology](/methodology/) - [30 questions about least privilege in the cloud](/faq/) - [The Access Debt Calculator: the data, downloadable](/datasets/) Last reviewed 16 Nov 2026. Canonical: https://leastprivilegeprinciple.com/terms/