Account Disabled ≠ Access Revoked: Why Identity and Authorization States Diverge
This article analyzes why disabling a user account in an identity platform doesn't automatically revoke access across connected systems, examining token lifecycles, session independence, and standards like NIST SP 800-63B and RFC 7009 to show how authorization states can persist after identity disablement.
The Problem: Account Disabled Does Not Mean Access Revoked
When an employee leaves, an administrator clicks "disable" in the unified identity platform. The account can no longer log in. But a previously approved connection between business systems may still read data. This gap is often dismissed with "the account is closed," yet real access flows through multiple stages: login to the identity provider, obtain a token, enter the application, and the application maintains its own session. Disabling occurs at one point; when other stages perceive that change depends on how systems propagate state.
The core question: what existing access does "can no longer log in" actually cover?
Account Disable First Affects the Next Login
The identity platform can block new logins and token issuance. However, the visitor may already hold a short-lived access token, or remain inside an application's own login session. NIST SP 800-63B-4 on Session Management explicitly separates identity-provider sessions from application sessions: ending one does not automatically end the other; access tokens and their associated refresh tokens may remain valid after the login session ends.
Microsoft's Entra ID emergency revocation documentation gives a concrete example: after an admin blocks new sign-ins and revokes refresh tokens, when existing access tokens expire and when the application's own session ends still depend on the application's access method and synchronization mechanism. This is product-specific implementation guidance, not a universal timeline, but it underscores a key principle: account status and actual accessible state are not the same table.
One Connection, Three Places That May Remain Valid
Consider a non-specific scenario: a business system was authorized to read a category of data from a data platform. After the personnel account is disabled, the front-door login is rejected, but the backend connection may still hold a token; the browser page of the business application may also maintain the application's own session. This "may" is an architectural deduction, not a claim that a specific real system has a vulnerability.
Observing this connection, the problem can be split into three moments: when the identity platform stops issuing new credentials; when already-issued tokens stop being accepted by the resource server; when the business application terminates its self-maintained session. If these three moments are not recorded in the same place, "disabled" only confirms the first step occurred.
OAuth 2.0 Token Revocation (RFC 7009) defines a revocation mechanism covering both refresh tokens and access tokens. The standard also notes that how quickly an access token can be invalidated depends on the token form and the resource server's validation method. This explains why "sending a revocation request" and "every entry point has rejected access" still require a confirmation step oriented to the actual access path.
Revocation Completion Should Be a Verifiable Business State
Many admin consoles excel at showing "account disabled" or "revocation request succeeded." But for the data that needs protection, the critical questions are: which application was authorized, which resources can it still access, when was the last successful access, and which system confirmed rejection of subsequent requests.
This does not demand every product draw a complex technical diagram. Instead, the manager-facing state can be simpler: new login blocked, existing token handling result, application session handling result, resource access verification result. The first three are configurations or actions; the last one approaches business effect. Separating "executed" from "verified" helps detect omissions and avoids reporting a still-propagating state as complete.
Google's OAuth documentation also lists user removal of third-party app authorization and application-initiated token revocation as separate operation entry points. This highlights another often-ignored boundary: beyond the personnel account, there exist authorization relationships granted to applications. When personnel leave, applications go offline, or services migrate, the governance object is not necessarily just a username.
From "Close One Account" to "Reclaim One Access Path"
For ordinary users, "disable" is best a clear result; for system builders, it is a chain of state changes that must close. Identity, tokens, application sessions, and resource access each have their own owners and their own time lags. Only by linking the authorization relationship with the actual access path can we know when a permission truly exits the business scene.
Next time you see "account disabled," the more valuable follow-up question may be: has the last link of this access path received and executed the revocation?
Sources and References
NIST SP 800-63B-4, Session Management: identity provider and application sessions are independent; access tokens may expire later than login sessions.
IETF RFC 7009, OAuth 2.0 Token Revocation: token revocation mechanism and revocation boundaries for different token forms.
Microsoft Learn, Revoke user access in an emergency in Microsoft Entra ID: differences in blocking new sign-ins, revoking refresh tokens, and existing access within that product.
Google for Developers, OAuth 2.0 for Client-side Web Applications: third-party application authorization removal and token revocation details.
Signed-in readers can open the original source through BestHub's protected redirect.
This article has been distilled and summarized from source material, then republished for learning and reference. If you believe it infringes your rights, please contactand we will review it promptly.
How this landed with the community
Was this worth your time?
0 Comments
Thoughtful readers leave field notes, pushback, and hard-won operational detail here.
