Griffin31
HEBook a demo
HE
Blog

Tokens in Microsoft Entra ID: what they are, how long they last, and how to protect them

1 November 2026 · Shlomi Fanan

Threats

Passwords and MFA protect the sign-in. Tokens are what’s left after it, and an attacker who gets them doesn’t need to sign in at all.

Two kinds of tokens

Microsoft splits tokens into two families. The difference decides what to protect first:

TokenWhat it’s forHow long it lastsCan it be revoked?
Primary Refresh Token (PRT)Requests tokens for any service, from a registered device90 days, renewed with every useYes
Refresh tokenRequests access tokens for one service90 days, renewed with every useYes
Access tokenGives access to the service itself60 to 90 minutes (75 on average)Only if the service supports CAE

Microsoft’s own conclusion: protect the sign-in session tokens (PRT and refresh tokens) first. They live for weeks or months, and an attacker who steals them stays in for a long time.

How long an access token really lasts

Worth knowing: you can no longer shorten refresh and session token lifetimes with a token lifetime policy. Microsoft retired that option in 2021. To force users to sign in again, use Conditional Access sign-in frequency.

What revoking sessions actually does

Revoking sessions and resetting the password invalidate the refresh token and the PRT. An access token that was already issued keeps working until it expires, unless the service supports CAE. So without CAE, an attacker can keep up to an hour and a half of access after you block them.

How to protect tokens: five layers

ControlWhat it doesLicense
Continuous access evaluation (CAE)Exchange, Teams and SharePoint reject a token as soon as a user is blocked, changes password or moves to a disallowed location. Critical events are enforced within minutes, location changes at onceEntra ID P1 (Conditional Access)
Token ProtectionBinds the session token to the device it was issued to. A token copied to another machine doesn’t workEntra ID P1
Compliant deviceA policy that requires an Intune-managed device. Even with a stolen token, the attacker’s device fails the checkEntra ID P1 and Intune
Phishing-resistant MFAPasskeys or FIDO2 through authentication strengths. A phishing page gets nothing it can reuseEntra ID P1
Risk-based policiesBlock or require re-authentication when the sign-in or user looks riskyEntra ID P2

Two more policies close common ways to get a token, both in P1: block the device code flow (we wrote about an attack like this in our article on device code phishing), and block authentication transfer, which moves a sign-in from a computer to a phone.

Where to start

  1. Make sure CAE is on and that no Conditional Access policy turns it off.
  2. Require a compliant device for company data, at least from Windows and macOS.
  3. Turn on Token Protection in report-only mode first, to see which apps don’t support it. Apple’s built-in Mail and Calendar apps, for example, don’t.
  4. Move to phishing-resistant MFA, admins and finance first.
  5. Block the device code flow and authentication transfer for everyone who doesn’t need them.
  6. Set sign-in frequency for sensitive apps and roles, instead of trying to shorten tokens.

Want to see how attackers actually steal tokens, and why passkeys alone aren’t enough? We covered it in our article on phishing, token theft and CAE.

How Griffin31 helps

Griffin31 checks the security settings in your Microsoft 365 tenant continuously, including your Conditional Access policies, and shows you what to fix first, with the evidence behind the results.

Want to know which token protections are already set up in your tenant? Griffin31 checks your Microsoft 365 security settings continuously and shows you what to fix first.

Sources