Step-up authentication
Asking a signed-in user to prove who they are again, usually with a second factor, before a sensitive action such as creating an API key or changing a role.
A session is only as good as the moment it began. Someone signed in with MFAMFA: Multi-factor authentication: signing in with a second proof, such as a one-time code from an authenticator app, as well as a password or SSO. this morning, then left the laptop open, or their session cookie was stolen by malware. Everything the session can do, the thief can do. Step-up narrows that: routine work goes on, but the few actions that would let an attacker stay in or widen their reach (minting keys, changing roles, editing secrets, turning MFA off) ask for a new proof first, if the last one is older than a few minutes.
NIST SP 800-63B-4 describes the idea in terms of assurance levels: an authentication process "may also provide step-up authentication to raise the session's AAL" (authenticator assurance level). On the API side, RFC 9470 standardises how a protected resource tells an OAuth client that the user's sign-in is not strong or recent enough: it answers with the error insufficient_user_authentication and can say which authentication context it wants (acr_values) or how recent the sign-in must be (max_age, in seconds).
Example
An admin opens Settings and clicks Create API key. Their last code was entered three hours ago, and the deployment requires one within 15 minutes for this action. The server refuses with a "step-up required" answer, the app asks for a code from the authenticator, sends it, and retries the action. A stolen cookie alone can browse, but cannot mint a key.
Common confusion
- Step-up is not MFA at sign-in. MFA proves who started the session; step-up asks again later, for one class of actions.
- It is not session expiry. A timeout ends the whole session; step-up keeps it and protects a few routes.
- API keys cannot answer a code. A system with step-up has to decide what a key may do on protected actions: refuse it, or allow only a narrowly scoped key meant for automation.
In L2M
Step-up is off unless STEP_UP_MAX_AGE_MINUTES is set. With it on, security-admin actions (users and roles, API keys and gateway clients, secrets, SSO mappings, notification and audit destinations, legal holds, retention and parts of data-subject requests) need an MFA check on the session within that many minutes; the studio asks for the code and retries. API keys are refused on those actions. The one exception is a service-account key with an admin ceiling and no project list, on an account that is still an admin, and it too is refused on actions that target a person's account. Each step-up is audited as mfa.step_up.