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.
Open the calculator ยท The field guide
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, and the boundary concept in detail is on 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 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, apply these rules to one identity type, and the glossary defines the terms used here.
Related pages
Sources
- AWS, permissions boundaries for IAM entities, read 16 Nov 2026
- AWS, service control policies, read 16 Nov 2026
- AWS, attribute-based access control for AWS, read 16 Nov 2026
- AWS, policy evaluation logic, read 16 Nov 2026
- AWS, managed policies for job functions, read 16 Nov 2026
- Microsoft, what is Azure role-based access control, read 16 Nov 2026
- Microsoft, what is Azure attribute-based access control, read 16 Nov 2026
- Microsoft, list Azure deny assignments, read 16 Nov 2026
- Microsoft, best practices for Azure RBAC, read 16 Nov 2026
- Google Cloud, overview of IAM Conditions, read 16 Nov 2026
- Google Cloud, deny policies, read 16 Nov 2026
- Google Cloud, Principal Access Boundary policies, read 16 Nov 2026
- Google Cloud, using IAM securely, read 16 Nov 2026
- AWS, testing IAM policies with the policy simulator, read 16 Nov 2026
Last reviewed 16 Nov 2026.