Explore our new Cybersecurity Terms Glossary, explained by our experts.
Cybersecurity Terms Glossary
Introspection asks the authorisation server whether a token is still valid. Revocation tells it to stop accepting one. Together they are how you end a session you no longer trust.

What Are Token Introspection and Revocation?
During an incident, somebody will ask how quickly you can cut an attacker’s access. If the answer is “when the tokens expire”, the two mechanisms in this entry are what you were missing.
Token introspection, defined in RFC 7662, is an endpoint where a resource server presents a token and asks the authorisation server what it knows about it. Token revocation, defined in RFC 7009, is an endpoint where a client tells the authorisation server to stop honouring a token it was issued.
A self-contained JSON Web Token is verified locally. The resource server checks a signature, reads the claims, and makes a decision without contacting anyone. That is why the architecture scales, and it is also why nothing you do at the authorisation server reaches a token already in circulation.
Disable the account and the token keeps working. Remove the user from every group and the token keeps working, because the scopes were baked in when it was issued. Change the password—still working. The token is a statement about a moment that has passed, and the resource server has no reason to doubt it until the expiry claim says otherwise.
Revoking a refresh token is more effective, because refresh exchanges necessarily contact the authorisation server. It stops the attacker obtaining new access tokens, which bounds the damage to the lifetime of the one they currently hold. This is why short access token lifetimes and real revocation are complements rather than alternatives.
An introspection request is a simple authenticated POST:
POST /oauth2/introspect HTTP/1.1
Host: login.example.co.uk
Content-Type: application/x-www-form-urlencoded
token=eyJhbGciOiJSUzI1NiIs...&token_type_hint=access_token
The response is JSON with an active boolean and, when active, the claims the authorisation server holds: scope, client identifier, subject, expiry. A false value means the token has expired, been revoked, or never existed—the specification deliberately does not distinguish, so that the endpoint cannot be used as an oracle.
Two things are worth being firm about. The endpoint must require authentication, or anyone can test stolen tokens against it. And caching is essential for anything high-volume: a call to the authorisation server on every API request reintroduces the central dependency that self-contained tokens were designed to remove. A cache measured in seconds gives most of the revocation benefit at a fraction of the cost.
You do not have to make one decision for everything. A workable pattern tiers it:
Opaque tokens, which carry no claims and are merely references, force introspection by design. They are slower and they give you immediate revocation for free, which is why financial-grade API profiles and the UK Open Banking specifications favour them for exactly the operations where that matters.
Revocation at the single-token level is rarely what an incident needs. What you actually want is “end every session for this user, everywhere, now”, and that is a design decision rather than a protocol feature.
The pieces that make it possible: a session identifier claim in issued tokens so sessions can be addressed as a group; a revocation list keyed on that identifier, checked by resource servers, with entries expiring when the underlying tokens would have; refresh token rotation with reuse detection, so that a replayed refresh token invalidates the whole chain and signals theft; and platform features where available—Microsoft Entra ID’s continuous access evaluation pushes revocation events to subscribed services within minutes rather than waiting for token expiry, and similar capabilities exist elsewhere under the OpenID Shared Signals framework.
Then test it. The most common finding in an incident retrospective is not that revocation was absent but that nobody had ever exercised it, the runbook referenced an administrative console that had changed, and the first attempt took forty minutes. Rehearse the full sequence—disable the account, revoke sessions and refresh tokens, confirm with introspection that the tokens are now inactive—and time it.
Finally, remember that revocation limits what a stolen token can do after you notice. Sender-constrained tokens reduce the value of theft before you notice, and conditional access reduces the chance of the theft succeeding. All three belong in the same design, alongside the issuance practices covered under OAuth 2.0.
At Privia Security, we test token lifecycle handling and session termination in our web application security services.
Because a self-contained access token is verified locally by the resource server, which reads the signature and claims without contacting the issuer. The account status is never consulted. Access continues until the token expires, which is why short access token lifetimes, refresh token revocation, and continuous access evaluation matter during an incident.
No, the cost is usually unjustified. Tier it instead: local verification for low-risk reads, introspection with a short cache for ordinary writes, and uncached introspection for high-value operations such as payments, bulk data export, or administrative changes. That gives near-immediate revocation where it matters without a round trip on every request.
Each refresh exchange issues a new refresh token and invalidates the previous one. If an old token is presented again, the authorisation server treats it as evidence of theft and revokes the entire token family, forcing re-authentication. It turns a replay attempt into both a containment action and a detection signal, and it is widely supported.
A mechanism where the identity provider pushes critical events—account disabled, password reset, risk detected, policy changed—to subscribed resource services so that sessions are terminated in near real time instead of at token expiry. Microsoft Entra ID implements it for Microsoft 365 services, and the OpenID Shared Signals framework generalises the pattern across providers.
You May Be Interested In These
We use a few cookies that are needed to make this website work, and they are always on. We would also like to set analytics cookies, to see which pages are useful; functional cookies, so that our live chat works; and advertising cookies, to measure our campaigns. We will not set any of those until you say yes, and you can change your mind at any time from the Cookie Settings link in the footer.
Choose the cookies you are happy for us to use. Strictly necessary cookies are always on; everything else is off until you turn it on.
These cookies keep the website working: they hold your security checks, your form submissions and the choice you make here. The site cannot run without them, so they are always on.
These let our live chat open and remember your conversation, and let the newsletter sign-up work. Switch them off and the chat window will not appear.
These tell us which pages are read and how people find us, so that we can improve the site. The reports are about groups of visitors, never about you personally.
These let us measure our advertising and show you our adverts on other websites, such as LinkedIn, Google and Meta. Switch them off and you will still see adverts, but they will be less relevant.