Skip to main content
Back
IntrusionMicrosoft Entra
4 min read

ConsentFix authorization code phishing through the Azure CLI app

Victim
Users in several Push Security customer environments (no account takeover confirmed)
When
December 2025, start not statedDisclosed Dec 11, 2025
Entry
An Azure CLI authorization code that the victim pasted into a compromised website after a real Microsoft sign-in.

Per Push Security, in December 2025 compromised websites reached through a search engine showed targeted business users a fake verification check, then opened Microsoft's real sign-in for the Azure CLI first-party app. After the sign-in, or with an existing session, the browser landed on a local address that held an authorization code, and the victim pasted that address into the attacker's page. Push blocked the pages before its customers interacted and reports no account takeover; in a lab run of the kit, the code was redeemed for tokens and non-interactive sign-ins from addresses in the United States and Indonesia followed.

What happened

External Parties
Compromised websites
Push Security
Internet
Microsoft Entra IDissues the code to Azure CLI
Microsoft Entra Tenant
Targeted user
Azure CLI app
Directory and Intune check-in

Replaying the ConsentFix December 2025 intrusion

  1. 1

    A gate on a trusted site

    Dec 2025, day not stated

    Compromised websites reached through a search engine show a fake verification check that lets only targeted business addresses go further.

  2. 2

    The code pasted back

    Dec 2025, day not stated

    The victim signs in on Microsoft's real page for the Azure CLI first-party app and pastes the local address that holds the authorization code into the attacker's page.

  3. 3

    Tokens from elsewhere

    Dec 2025, day not stated

    The attacker redeems the code for tokens with no client secret; in a lab run of the kit, non-interactive sign-ins from the United States and Indonesia follow against the directory and Intune check-in.

Impact

Push Security blocked the pages before its customers interacted, and no account takeover is reported; in a lab run, the kit turned a pasted code into tokens used from addresses in the United States and Indonesia against the directory and Intune check-in.

Outcome

Where Microsoft Entra controls fit

Microsoft documents that disabling sign-in on a service principal stops every token for that application; its page does not name Azure CLI, the setting would end all legitimate CLI use, and Microsoft does not recommend it for a first-party application. Assignment required, and Conditional Access blocks on the directory and Azure management resources, would have narrowed that sign-in to the administrators and developers who use the CLI, though the sources do not say whether the targeted users were among them. A Conditional Access policy that requires the compliant network would have refused new tokens requested from elsewhere, though not the first redemption of the code. The fake check on compromised websites sits outside every cloud control. Check which of your users can sign in to Azure CLI and the other first-party clients.

Could this have been prevented?

In hindsight, no control would have stopped it. 4 controls would have limited how far it went. 1 step has no cloud-provider control. The counts below are control assessments, not attack stages: several can address the same step. The same replay with the controls in place before the incident shows where each phase would have been refused or bounded.

0
Prevented
4
Contained
1
Not covered

How it could have been prevented

The same replay with the controls in place. Green marks where a step is refused, blue where it is bounded, and red where no control reaches it. Open a row to read the control, why it acts, and its caveat.

External Parties
Compromised websites
Push Security
Internet
Microsoft Entra IDissues the code to Azure CLI
Microsoft Entra Tenant
Targeted user
Azure CLI app
Directory and Intune check-in

Replaying the ConsentFix December 2025 intrusion

1 · A gate on a trusted site

1 Not covered

2 · The code pasted back

3 Contained

3 · Tokens from elsewhere

1 Contained

Key facts

  • Push Security detected the campaign in December 2025, day not stated, on compromised websites in several customer environments and blocked it before customers interacted
  • The fake verification check let only business email addresses on the attacker's target list go further and sent everyone else back to the original site
  • The sign-in was Microsoft's real page for the Azure CLI first-party app, which Push says cannot be deleted or blocked; it ended on a local address that held an authorization code
  • Azure CLI is a public client, so the code was redeemed for tokens with no client secret; glueckkanja quotes Microsoft that codes expire after about 10 minutes
  • In a lab run of the kit, non-interactive sign-ins from addresses in the United States and Indonesia followed against the directory and Microsoft Intune check-in resources
  • glueckkanja listed 11 first-party apps vulnerable to the technique on December 31, 2025, and Push's January 2026 debrief says the campaign appears to be linked to APT29

Sources

Every fact in this breakdown comes from the public reporting below. The replay is a model of that reporting, not a reconstruction of the environment it describes.

  1. 1.
    ConsentFix: Analyzing a browser-native ClickFix-style attack that hijacks OAuth consent grantsPush Security · Dec 11, 2025 · ResearchDelivery through compromised sites found by search, the email gate, the real Microsoft sign-in to the Azure CLI app, the local redirect with the code, the paste, the code redemption, the non-interactive sign-ins that followed in a lab run, blocking before customer interaction
  2. 2.
    ConsentFix debrief: latest community insights, recommendations, and predictionsPush Security · Jan 14, 2026 · ResearchScale across compromised sites and customer environments, other vulnerable first-party apps, the attribution hedge, recommended restrictions
  3. 3.
    AuthCodeFix aka ConsentFixglueckkanja · Dec 31, 2025 · ResearchThe requested client and resource, access and refresh tokens from the code, the code lifetime, pre-consented first-party apps, user assignment, Conditional Access blocks, Token Protection, the compliant network
  4. 4.
    ConsentFix (a.k.a. AuthCodeFix): Detecting OAuth2 Authorization Code PhishingNVISO · Jan 29, 2026 · ResearchRedemption at the token endpoint with no client secret, the sign-in log pattern, user assignment on first-party apps

The estate in the replay is illustrative and drawn at the level of attack stages; the victim's real environment is not public. Verdicts describe what a control does when it is in place before the incident. They are not an assessment of the victim's security program.

More cloud attack breakdowns

All breakdowns