How the attack works
Before an application can read a user's mail or files, someone has to consent to the permissions it asks for. The tenant's authorization policy decides whether ordinary users may give that consent themselves. The setting can be off, limited to applications from verified publishers and to permissions classified as low impact (microsoft-user-default-low), managed by Microsoft (microsoft-user-default-recommended, the default for a new tenant), or open for any application and any permission that does not need admin consent (microsoft-user-default-legacy).
Organizations restrict the setting to stop consent phishing, in which a user is tricked into granting a malicious application access. An attacker who holds the Privileged Role Administrator or Global Administrator role, or an application with Policy.ReadWrite.Authorization, can quietly open it again. The change gives no access by itself. It prepares the next step: from then on one user who accepts a prompt hands an application a lasting token to their data. The same policy holds allowUserConsentForRiskyApps, which lets users consent to requests that Microsoft considers risky.
Microsoft Entra ID records the change as Update authorization policy in the AuthorizationPolicy category of the Core Directory service. The assigned consent policy is stored as ManagePermissionGrantsForSelf.<policy id>, and the consent rule on this page looks for that value in the modified properties of the event. The same activity is written for every other change to the authorization policy, such as who may invite guests or whether users can register applications, and the broader rule on this page covers those. What a consent policy allows is written in the app consent policy object itself, and a third rule reads additions, updates and deletions of those objects (Add permission grant policy, Update permission grant policy, Delete permission grant policy). A custom policy that is already assigned can be widened that way without any change to the authorization policy.
What an attacker needs
- A compromised identity with the Privileged Role Administrator or Global Administrator role, or an application with the Microsoft Graph permission
Policy.ReadWrite.Authorization.
Impact
Users can again consent to applications, with the legacy policy to any application for any permission that does not need admin consent. Consent phishing works again, and each consent gives an application lasting access to the mail, files or profile of the user who accepted it. Consent grants that already exist are not touched by a change of the setting in either direction.
Detection rules
App Consent Policy Added, Updated or Deleted
Detects a change to an app consent policy, the object that describes which consent requests are allowed. The authorization policy only points to a consent policy. What users may consent to is written in the policy the pointer names, as condition sets that include or exclude applications and permissions. An attacker who can write these policies can therefore widen user consent without touching the authorization policy: by adding an include condition set to a custom policy that is already assigned, or by creating a permissive policy and using it in a custom role. Microsoft lists the three activities in the PermissionGrantPolicy category of the Core Directory service and documents their names only. The rule matches the names, and the policy with its old and new values is in TargetResources.
title: App Consent Policy Added, Updated or Deleted
id: e98cab8f-b36e-4f5a-a7a0-7e96c0910071
status: experimental
description: >-
Detects a change to an app consent policy, the object that describes which
consent requests are allowed. The authorization policy only points to a
consent policy. What users may consent to is written in the policy the
pointer names, as condition sets that include or exclude applications and
permissions. An attacker who can write these policies can therefore widen
user consent without touching the authorization policy: by adding an
include condition set to a custom policy that is already assigned, or by
creating a permissive policy and using it in a custom role. Microsoft lists
the three activities in the PermissionGrantPolicy category of the Core
Directory service and documents their names only. The rule matches the
names, and the policy with its old and new values is in TargetResources.
references:
- https://learn.microsoft.com/en-us/entra/identity/monitoring-health/reference-audit-activities
- https://learn.microsoft.com/en-us/entra/identity/enterprise-apps/manage-app-consent-policies
- https://attack.mitre.org/techniques/T1484/
author: Forestall
date: 2026-11-18
tags:
- attack.defense-impairment
- attack.privilege-escalation
- attack.t1484
logsource:
product: azure
service: auditlogs
detection:
selection:
OperationName:
- Add permission grant policy
- Update permission grant policy
- Delete permission grant policy
filter_failed:
Result: failure
condition: selection and not filter_failed
falsepositives:
- Administrators who build or adjust a custom app consent policy as part of a consent governance project. These changes are rare, so match each one to a change record.
level: medium
False positives
- Administrators who build or adjust a custom app consent policy as part of a consent governance project. These changes are rare, so match each one to a change record.
Tenant Authorization Policy Updated
Detects every successful update of the tenant's authorization policy, the audit activity "Update authorization policy" in the AuthorizationPolicy category of the Core Directory service. The policy is a singleton that holds the default permissions of all users: whether users may consent to applications and to requests Microsoft considers risky, who may invite guests, which role guests receive, and whether users can register applications, create tenants or read the BitLocker keys of their own devices. It changes rarely, so each event is worth reading. The rule does not say which property changed. Compare the old and the new values in TargetResources.
title: Tenant Authorization Policy Updated
id: 9f4cf9d2-b563-4f0b-a506-e947b7b8aabb
status: experimental
description: >-
Detects every successful update of the tenant's authorization policy, the
audit activity "Update authorization policy" in the AuthorizationPolicy
category of the Core Directory service. The policy is a singleton that
holds the default permissions of all users: whether users may consent to
applications and to requests Microsoft considers risky, who may invite
guests, which role guests receive, and whether users can register
applications, create tenants or read the BitLocker keys of their own
devices. It changes rarely, so each event is worth reading. The rule does
not say which property changed. Compare the old and the new values in
TargetResources.
references:
- https://learn.microsoft.com/en-us/entra/identity/monitoring-health/reference-audit-activities
- https://learn.microsoft.com/en-us/graph/api/resources/authorizationpolicy
- https://learn.microsoft.com/en-us/graph/api/resources/defaultuserrolepermissions
- https://attack.mitre.org/techniques/T1484/
author: Forestall
date: 2026-11-16
tags:
- attack.defense-impairment
- attack.privilege-escalation
- attack.t1484
logsource:
product: azure
service: auditlogs
detection:
selection:
OperationName: Update authorization policy
filter_failed:
Result: failure
condition: selection and not filter_failed
falsepositives:
- Planned changes to user settings, external collaboration settings or device settings by the identity team. Each one should have a change record. A change that tightens a setting needs no action.
level: low
False positives
- Planned changes to user settings, external collaboration settings or device settings by the identity team. Each one should have a change record. A change that tightens a setting needs no action.
User Consent Setting for Applications Changed
Detects an update of the tenant's authorization policy whose modified properties carry a user consent policy. Microsoft stores the policy that lets users consent to applications as ManagePermissionGrantsForSelf followed by a policy ID, for example microsoft-user-default-legacy (consent for any application), microsoft-user-default-recommended (managed by Microsoft) or microsoft-user-default-low (verified publishers, low impact permissions). The value appears as the old or the new value, so the rule fires when the setting is weakened and when it is tightened. Read both values before escalating. Microsoft does not document the layout of this audit event. The companion rule on every update of the authorization policy covers a tenant that records the change without the policy value.
title: User Consent Setting for Applications Changed
id: 4fd1775f-08ba-4aa8-8bfe-44b4d8824b77
status: experimental
description: >-
Detects an update of the tenant's authorization policy whose modified
properties carry a user consent policy. Microsoft stores the policy that
lets users consent to applications as ManagePermissionGrantsForSelf followed
by a policy ID, for example microsoft-user-default-legacy (consent for any
application), microsoft-user-default-recommended (managed by Microsoft) or
microsoft-user-default-low (verified publishers, low impact permissions).
The value appears as the old or the new value, so the rule fires when the
setting is weakened and when it is tightened. Read both values before
escalating. Microsoft does not document the layout of this audit event. The
companion rule on every update of the authorization policy covers a tenant
that records the change without the policy value.
references:
- https://learn.microsoft.com/en-us/entra/identity/enterprise-apps/configure-user-consent
- https://learn.microsoft.com/en-us/graph/api/resources/defaultuserrolepermissions
- https://learn.microsoft.com/en-us/entra/identity/monitoring-health/reference-audit-activities
- https://attack.mitre.org/techniques/T1484/
author: Forestall
date: 2026-11-16
tags:
- attack.defense-impairment
- attack.privilege-escalation
- attack.t1484
logsource:
product: azure
service: auditlogs
detection:
selection:
OperationName: Update authorization policy
TargetResources|contains: ManagePermissionGrantsForSelf
filter_failed:
Result: failure
condition: selection and not filter_failed
falsepositives:
- Planned hardening, for example the move from consent for any application to the Microsoft managed policy or to verified publishers only. Match the event to the change record and read the new value.
- An update of another property of the authorization policy, if the tenant writes unchanged properties into the event as well. Compare the old and the new value of the consent policy.
level: medium
False positives
- Planned hardening, for example the move from consent for any application to the Microsoft managed policy or to verified publishers only. Match the event to the change record and read the new value.
- An update of another property of the authorization policy, if the tenant writes unchanged properties into the event as well. Compare the old and the new value of the consent policy.
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.
Consent setting opened, consent given, setting closed again
Opening user consent is only a preparation. An attacker who opens the setting, has a user or their own account consent to an application and then restores the setting leaves a tenant that looks as strict as before, with one more application holding access.
- Audit log:
Update authorization policywithManagePermissionGrantsForSelfin the modified properties ofTargetResources. Keep the old and the new value. - Within 24 hours: one or more successful
Consent to applicationevents. - Within the same 24 hours: a second
Update authorization policywhose new value equals the old value of the first. - Review every consent between the two changes, and compare each with what the restored setting would have allowed.
EventsUpdate authorization policyConsent to application
LimitsAn administrator who loosens the setting for a planned rollout and tightens it afterwards produces the same sequence. The 24 hours are a starting point. Consent given while the setting stays open is not separated from ordinary consent by this hypothesis.
Triage
- Read the old and the new value in TargetResources. A new value with
microsoft-user-default-legacyallows consent for any application, and a value without anyManagePermissionGrantsForSelfentry turns user consent off. - Identify the actor (InitiatedBy) and look for a change record. This setting changes rarely, and a change by an application or outside a planned project needs an explanation.
- Check the actor's sign-ins before the change for unfamiliar IP addresses or risk.
- Search the audit log after the change for
Consent to applicationevents by users, and review the applications and the permissions involved. - Look at what else the same actor changed in the session, such as other properties of the authorization policy or role assignments.
Response
Once the attack is confirmed:
- Set the user consent setting back to the intended value. It applies to consent requests from that moment on.
- Review every
Consent to applicationevent since the setting was weakened. Remove the grants of applications you do not trust and disable their service principals. - Revoke the sessions and reset the credentials of the identity that made the change, and remove its Privileged Role Administrator or Global Administrator role until the investigation is closed.
- For each user who consented to a suspicious application, review what the application could read and treat that data as exposed.
- Compare the other properties of the authorization policy with your record, because the same actor may have changed them as well.
Mitigation
- Limit user consent to applications from verified publishers and to permissions you classified as low impact, let Microsoft manage the setting, or turn it off. Enable the admin consent workflow so that users can request what they cannot grant.
- Keep
allowUserConsentForRiskyAppsset to false, as Microsoft recommends. - Limit Privileged Role Administrator and Global Administrator to a few accounts managed with Privileged Identity Management, and review the applications that hold
Policy.ReadWrite.Authorization. - Record the intended values of the authorization policy and compare them regularly, so that drift is found even when an alert is missed.
Frequently asked questions
Does restoring the setting remove the consents users gave in the meantime?
No. Microsoft documents that updates to the user consent settings affect only future consent operations and that existing grants remain unchanged. After restoring the setting, review the consents given while it was weakened and remove the grants of applications you do not trust.
How do I see from the event whether the setting was weakened or tightened?
Compare the old and the new value in TargetResources. The assigned policy is stored as ManagePermissionGrantsForSelf followed by a policy ID. The ID microsoft-user-default-legacy allows consent for any application, microsoft-user-default-recommended is the policy Microsoft manages, microsoft-user-default-low limits consent to verified publishers and low impact permissions, and no such entry means user consent is off. The rule matches a change in either direction.
Why is there a second rule for every update of the authorization policy?
Microsoft documents the activity name and the format of the consent policy value, but not the layout of this audit event. If a tenant records the change without the policy value, the consent rule stays silent and the broader rule still fires. The broader rule also covers the other settings of the same policy, such as who may invite guests and whether users can register applications.
Does this page detect the malicious consent itself?
No. It detects the change of the tenant setting that makes user consent possible. The consent of a user to an application is a separate audit event, Consent to application, and is covered by the page Illicit Consent Grant.
References
- MITRE ATT&CK T1484: Domain or Tenant Policy Modification
- Configure how users consent to applications
- Manage app consent policies (built-in policies and Microsoft recommended settings)
- Microsoft Entra audit log activities (Core Directory, AuthorizationPolicy)
- authorizationPolicy resource type (Microsoft Graph)
- defaultUserRolePermissions resource type (permissionGrantPoliciesAssigned)
- Microsoft Entra security operations guide for applications (end-user consent)