Entra ID Sign-ins
AADSTS result codes and risk detections, read as investigative signals.
Microsoft's AADSTS reference is written for developers fixing an app. Reading a tenant's sign-in log after an incident is a different job. Thirty 50126 failures against thirty users is a password spray, not thirty typos.
Before you read the table
- Successful sign-ins carry risk detections too. A success flagged atypical travel is more urgent than any failure on the page.
- The authentication requirement field proves MFA. A policy existing does not. singleFactorAuthentication means MFA was not satisfied.
26 rows.
| Code | Meaning | Investigative use |
|---|---|---|
| 50126 | Error validating credentials due to invalid username or password. | The workhorse failure code, and the one password spraying is made of. A single occurrence is a typo; the same code against many DIFFERENT user principal names from one IP or one ASN, a few attempts per account, is a spray. |
| 50053 | Account is locked because the user tried to sign in too many times with an incorrect ID or password (smart lockout). | Proof that the volume against a single account crossed Entra's own threshold. Useful in two directions: it confirms a brute-force attempt, and it explains a wave of user-reported lockouts that would otherwise look like an outage. |
| 50034 | The user account does not exist in the directory. | User enumeration. An attacker is testing a list of guessed or harvested addresses to learn which ones are real before spending attempts on passwords. The valid names are the ones that come back with 50126 instead. |
| 50057 | User account is disabled. | Usually a stale credential list: an attacker working from a breach dump or an old export that still contains leavers. Tells you the source data is old, which is a useful attribution detail. |
| 50055 | The password is expired. | Notable only because it means the password supplied was OTHERWISE CORRECT. In a spray investigation this is a hit, not a miss. The credential is valid and the attacker now knows it. |
| 50133 | Session is invalid due to expiration or recent password change. | Expected noise after a password reset, including the reset YOU performed during containment. Useful for confirming that a revocation actually took effect. |
| 500121 | Authentication failed during strong authentication request. | The MFA-fatigue code. It covers a denied push, a timed-out prompt, and a user who simply did not respond, so a repeated run of it against one account means someone holds a valid password and is pushing prompts at a person who keeps refusing. The dangerous moment is the FIRST success after a run of these: that is the prompt the user finally approved. |
| 50074 | User did not pass the MFA challenge (strong authentication required). | The password was accepted and MFA stopped the sign-in. Every occurrence is a credential you should now consider known to someone else: the control worked, but the secret is still burned. |
| 50076 | Due to a configuration change made by the administrator, or because the user moved to a new location, the user must use multi-factor authentication. | Routine when a Conditional Access policy fires on a new network. Suspicious when the "new location" is somewhere the user has never been and never travels to. |
| 50079 | The user is required to enroll in multi-factor authentication (proof-up required). | Dangerous when it appears on an account that should ALREADY have MFA. An attacker with a valid password reaching an un-enrolled account gets to register their own authenticator, which converts a stolen password into durable access that survives a reset. |
| 53003 | Access has been blocked by Conditional Access policies. | A control fired. The question is always which policy and why: the sign-in log names the policy, so this is a rare case where the log tells you exactly what saved you. A run of these from one identity is someone probing for a path the policy set does not cover. |
| 53000 | Device is not compliant; the user must enroll or remediate. | A correct password arriving from an unmanaged device. Legitimate for a user on a new laptop, and exactly what a credential thief looks like. The distinguishing evidence is elsewhere (location, user agent, whether the real user was working). |
| 53001 | Device is not domain joined. | Same reading as 53000. The credential was accepted, the endpoint was not. |
| 53002 | The application used is not an approved application for Conditional Access. | Frequently an attacker choosing a legacy or non-standard client on purpose, because approved-app policies are one of the controls that legacy protocols bypass. |
| 50097 | Device authentication is required. | Common in hybrid-join environments. Read alongside 53000/53001 rather than on its own. |
| 50158 | External security challenge was not satisfied. | A third-party or external MFA/challenge provider rejected the sign-in, so the corroborating evidence lives in that provider's logs, not in Entra's. |
| 65001 | The user or administrator has not consented to use the application. | The footprint of an illicit consent-grant attempt. A user was sent to a real Microsoft consent screen for an attacker-registered application; the code means they had not (yet) agreed. The follow-up question is whether a later attempt succeeded: a granted consent produces no failure code at all, only an audit-log entry. |
| 90094 | The grant requires administrator permissions. | An application asked for permissions only an admin can grant. Mailbox-wide or directory-wide scopes. Legitimate for enterprise software rollouts, and precisely what a serious consent-phishing attempt looks like. Read the requested scopes, not the app's display name, which the attacker chose. |
| 7000215 | Invalid client secret provided. | A workload identity failing to authenticate. Routine after a secret rotation and alarming otherwise: someone is trying secrets against a service principal, and service principals frequently hold broader permissions than any human account and are excluded from MFA by nature. |
| 700016 | Application with the given identifier was not found in the directory. | Often an attacker probing for the presence of a well-known first-party or third-party app ID in the tenant, or a stale integration. Read as reconnaissance when it comes in a burst of distinct app IDs. |
| 50105 | The signed-in user is not assigned to a role for the application. | Authentication succeeded; authorization did not. The identity is valid, which makes this a confirmation of compromise rather than a failed attempt, if the user had no reason to touch that app. |
| 50173 | Fresh authentication token is required. The password has changed. | Confirms a password change invalidated an existing session. During containment this is the code that proves your reset actually took effect. |
| 50058 | A silent sign-in request was sent but no user is signed in. | Ordinary browser/SSO behavior. Almost never meaningful on its own; useful only for reconstructing a session's shape. |
| 50140 | The sign-in was interrupted by the "Keep me signed in" prompt. | Interactive-session furniture. Its value is confirming a human was at a browser rather than a script. |
| 16000 | User account selection is required. The session has multiple accounts. | Noise in shared or multi-tenant browsers; not a security signal. |
| 90002 | Tenant not found. | Someone is authenticating against a tenant identifier that does not exist here, typically a misdirected integration, occasionally reconnaissance across tenant names. |