Permission
A permission is the ability to perform a specific action on a specific resource, such as reading a file or editing a record. In everyday terms it reflects the general idea of allowing someone to do something, but in identity systems it is a concrete, granular grant rather than a vague consent. Permissions define what a principal is allowed to do once their identity has been verified.
A permission is a discrete authorization primitive that couples an allowed operation (for example, read, write, delete) with a target object or resource, defining what a principal may do after identification and authentication have completed. Permissions are enforced at the authorization stage and are typically aggregated into higher-level constructs depending on the access control model in use: in RBAC they are assigned to roles which are then granted to subjects, while in ABAC or PBAC the effective set of permitted actions is derived at evaluation time from attributes and policy rather than statically pre-bound. Note that a permission expresses only the granted capability; the runtime decision to allow or deny a given request is made by a policy decision point and applied by a policy enforcement point, and the terminology and granularity of permissions vary by vendor and deployment context.
Why it matters
Permissions are the atomic unit of authorization, and the correctness of an entire access control system ultimately rests on whether the right principals hold the right permissions on the right resources. When permissions accumulate beyond what a principal actually needs, the result is over-entitlement and privilege creep, which expand the blast radius of any compromised account. Because permissions define concrete capabilities rather than vague consent, they are the primary object of access reviews, certifications, and least-privilege analysis in identity governance.
The granularity and clarity of permissions directly affect both security posture and auditability. A permission model that is too coarse forces administrators to grant more than necessary, while one that is excessively fine-grained can become unmanageable and error-prone. In most deployments, permissions are not assigned directly to individuals but aggregated into higher-level constructs such as roles or derived at evaluation time from policy, which shifts the challenge from granting individual rights to reasoning about how those aggregations resolve to effective access.
It is worth emphasizing that a permission expresses only a granted capability; it does not by itself make an allow-or-deny decision. Conflating the existence of a permission with the runtime enforcement of it is a common source of confusion, and understanding the separation between the two is essential for anyone auditing why a particular request succeeded or failed.
Who it's relevant to
Inside Permission
Common questions
Answers to the questions practitioners most commonly ask about Permission.
