1. Resources
  2. Detection rules
  3. Federated Identity Credential Added to an Application

Federated Identity Credential Added to an Application detection rules

An attacker adds a federated identity credential to an app registration, so tokens from an external identity provider they control are exchanged for tokens of the application. No secret is created and nothing expires.

Platform
Microsoft Entra ID
Severity
High
Log source
Microsoft Entra ID audit logs
Last updated
Rule formats
Sigmaand queries for: SplunkMicrosoft SentinelIBM QRadar
What an attacker needs
An owner or administrator role that can manage the application's credentials, and an external identity provider whose tokens the attacker can obtain.
Impact
App-only access as the application from outside the organization, with no secret to steal or rotate, that lasts until the federated credential is found and deleted.
Response
Delete the federated identity credential from the app registration, and remove any other credentials and owners on the application that nobody can explain.
Mitigation
Limit who owns or administers applications, and include federated identity credentials in the reviews and alerts you already have for client secrets and certificates.
Audit activity
Update application
Threat groups
Storm-0501

How the attack works

Workload identity federation lets an application in Microsoft Entra ID trust tokens from an external identity provider instead of holding a client secret or a certificate. The trust is a federated identity credential on the app registration. It names an issuer, a subject and an audience, and Microsoft Entra ID exchanges an external token that matches all three for an access token of the application. It is meant for workloads such as deployment pipelines, Kubernetes clusters and software running in other clouds.

Whoever can manage the credentials of an application can add one: the owners of the application and the Application Administrator, Cloud Application Administrator, Hybrid Identity Administrator and Global Administrator roles. An attacker who controls such an identity adds a credential whose issuer and subject point to a workload they run, for example their own repository or an identity provider they host. From then on they obtain tokens as the application from outside the tenant, with every directory role and API permission the application holds.

The result is the same as adding a client secret, but it is easier to miss. No secret value exists, the credential has no expiry date that would bring it up for review, and monitoring built for new secrets and certificates does not see it. Microsoft's audit activity reference has no activity for federated identity credentials. Public detection content reports that the change is logged by Core Directory as Update application with FederatedIdentityCredentials among the modified properties. A federated credential on a user-assigned managed identity is a separate case that is recorded in the Azure Activity log.

What an attacker needs

  • Control of an identity that can manage the application's credentials, which means an owner of the application, one of the administrator roles above, or an application with a Microsoft Graph permission such as Application.ReadWrite.All.
  • An external identity provider, or a workload at a public one, whose tokens the attacker can obtain.
  • A target application with valuable roles or permissions.

Impact

App-only access as the application, from outside the organization and without any secret to steal or rotate. The access survives password resets of the account the attacker started from and the rotation of the application's secrets and certificates, and it lasts until the federated credential is found and deleted. When the application holds a privileged directory role or a high-impact Microsoft Graph permission, this is a path to control of the tenant.

Detection rules

Federated Identity Credential Added or Changed on an Application

Detects an Update application event whose modified properties include FederatedIdentityCredentials, which means a federated identity credential of the app registration was added, changed or removed. The activity name is in Microsoft's audit activity reference. The property name is not: it comes from public detection content, so confirm the shape of the event in your own tenant before relying on the rule. Read the issuer and the subject of the credential in the event, because they decide who can obtain tokens as the application.

LevelMediumStatusExperimentalLog sourceMicrosoft Entra ID audit logs
title: Federated Identity Credential Added or Changed on an Application
id: 419fd417-5a82-4ce4-a139-d6af90b0669e
status: experimental
description: >-
  Detects an Update application event whose modified properties include
  FederatedIdentityCredentials, which means a federated identity credential
  of the app registration was added, changed or removed. The activity name
  is in Microsoft's audit activity reference. The property name is not: it
  comes from public detection content, so confirm the shape of the event in
  your own tenant before relying on the rule. Read the issuer and the subject
  of the credential in the event, because they decide who can obtain tokens
  as the application.
references:
  - https://learn.microsoft.com/en-us/entra/identity/monitoring-health/reference-audit-activities
  - https://learn.microsoft.com/en-us/entra/workload-id/workload-identity-federation-create-trust
  - https://github.com/elastic/detection-rules/blob/main/rules/integrations/azure/persistence_entra_id_federated_identity_credential_modified.toml
  - https://attack.mitre.org/techniques/T1098/001/
author: Forestall
date: 2026-11-14
tags:
  - attack.persistence
  - attack.privilege-escalation
  - attack.t1098.001
logsource:
  product: azure
  service: auditlogs
detection:
  selection:
    OperationName: Update application
    TargetResources|contains: FederatedIdentityCredentials
  filter_failed:
    Result: failure
  condition: selection and not filter_failed
falsepositives:
  - Platform and deployment teams that set up workload identity federation for pipelines, Kubernetes clusters or workloads in other clouds. Filter their identities in InitiatedBy once they are documented and review the issuer of each new credential.
  - Removal of a federated identity credential, which changes the same property. The old and the new value in the event show whether a credential was added or removed.
level: medium

False positives

  • Platform and deployment teams that set up workload identity federation for pipelines, Kubernetes clusters or workloads in other clouds. Filter their identities in InitiatedBy once they are documented and review the issuer of each new credential.
  • Removal of a federated identity credential, which changes the same property. The old and the new value in the event show whether a credential was added or removed.

Detection logic

Patterns that need more than one event, a baseline or knowledge of your own environment. Each is a hypothesis to build and tune in your SIEM, not a ready rule.

First token exchange with a new federated identity credential

Detection hypothesisSequence of eventsJoins two logsNeeds a baseline

A federated identity credential is used when a token of its issuer is exchanged for a token of the application, and the service principal sign-in log names the federated credential that was used. A credential that appears in sign-ins for the first time after the change, from a network the application's workloads do not use, is the attacker's external workload signing in.

  1. Audit log: Update application with FederatedIdentityCredentials in TargetResources, as the rule on this page reads it. Take the application and the time.
  2. Service principal sign-in log (AADServicePrincipalSignInLogs in Microsoft Sentinel), 30 days before the change: the FederatedCredentialId and IPAddress values of the application's AppId.
  3. Same log, 7 days after the change: sign-ins with a FederatedCredentialId not seen before. Read IPAddress, AutonomousSystemNumber and ResourceDisplayName.
  4. Alert, and compare the network with the platform of the issuer the credential names, such as your own CI system or cloud account.

EventsUpdate application

LimitsA pipeline or cluster that the platform team federates on purpose matches as well, so confirm the change first. Workloads at large public platforms sign in from shared address ranges, which makes the network a weak signal there. The service principal sign-in log is a separate diagnostic category. The 30 and 7 days are starting points.

Triage

  • Identify the application (TargetResources) and the actor (InitiatedBy). A credential added by a person outside a deployment or platform team needs an explanation.
  • Read the issuer, subject and audience of the new credential in the modified properties or on the app registration. Do they belong to a repository, cluster or cloud account of the organization?
  • Check the roles and API permissions of the application to understand what the credential gives access to.
  • Review the service principal sign-ins of the application after the change for new IP addresses and unusual API activity.
  • If the credential is not explained, delete it, review what the application did since it was added and investigate the actor's other changes.

Response

Once the attack is confirmed:

  1. Delete the federated identity credential from the app registration. As long as it exists, a token from the external issuer is exchanged for a token of the application.
  2. Check the application for other things the same actor could have added (further federated credentials, client secrets, certificates and owners) and remove everything that nobody can explain.
  3. Revoke the sessions and reset the credentials of the identity that added the credential, and remove its ownership of the application or its administrator role.
  4. Review what the application did since the credential was added. Read its service principal sign-ins and the audit log for role assignments, permission grants and credentials it created, and undo them.
  5. Go through the federated identity credentials of the other applications the same actor could manage, and delete those that point to an issuer or subject nobody recognizes.

Mitigation

  • Keep privileged directory roles and high-impact Microsoft Graph application permissions off applications wherever possible.
  • Limit who owns or administers applications, and review the owners of privileged applications regularly.
  • Inventory the federated identity credentials of your applications and check each issuer and subject against a workload you know. Delete credentials whose workload no longer exists.
  • Include federated identity credentials in the reviews and alerts you already have for client secrets and certificates.
  • Review the sign-ins of service principals for token requests that come from unexpected networks.

Frequently asked questions

Why do alerts for new client secrets miss a federated identity credential?

A federated identity credential creates no secret value and no certificate, and it has no expiry date. Monitoring built for new secrets and certificates therefore does not see it. This rule looks for an `Update application` event whose modified properties include `FederatedIdentityCredentials`.

Is the property name the rule matches documented by Microsoft?

Only in part. The activity name `Update application` is in Microsoft's audit activity reference. The property name `FederatedIdentityCredentials` is not, because it comes from public detection content. Confirm the shape of the event in your own tenant before you rely on the rule.

Does the rule cover federated credentials on managed identities?

No. A federated credential on a user-assigned managed identity is a separate case that is recorded in the Azure Activity log, not in the Microsoft Entra audit log this rule reads. Monitor that log as well if you use such identities.

Does rotating the application's secrets remove the attacker's access?

No. The federated credential is independent of the client secrets and certificates of the application, so rotating them changes nothing. The access also survives a password reset of the account the attacker started from. It ends when the federated credential itself is deleted.

References

We respect your privacy

We use cookies to keep this site secure and working properly. With your permission, we also use optional cookies to understand usage and improve the experience. Cookie Policy

You can change your choice at any time.