# OAuth Consent Phishing: Password-less Account Takeover via Malicious App Grants (Expel topic; EvilTokens, ConsentFix, FBI PSA)

> Consent phishing tricks a user into approving an attacker-controlled OAuth application on a genuine identity-provider consent screen, giving the attacker a persistent token without ever stealing a password. Access survives password resets and MFA until the grant is revoked.

- **Published:** 2026-11-16T20:21:37Z
- **Last reviewed:** 2026-11-16T20:21:37Z
- **Canonical:** https://intel.threadlinqs.com/threat/TL-2026-6177
- **Card image:** https://intel.threadlinqs.com/og/threat/TL-2026-6177.png
- **ID:** TL-2026-6177
- **Severity:** HIGH
- **Category:** PHISHING
- **Status:** TRACKING
- **Detections:** 9 (1 shown free below) · **IOCs:** 10 (2 shown free below)

## Description

SOURCING NOTE: the Expel blog post that triggered this record ('Consent phishing skips the password and walks in the front door') could not be retrieved; WebFetch returned only the title and WebSearch did not surface it. Nothing below is attributed to Expel. This record is a topic-level synthesis from other fetched sources (Push Security, Cloud Security Alliance, a secondary summary of an FBI IC3 PSA). Severity is an analyst assessment based on the reported scale of related campaigns, not a figure taken from the Expel article.

Attack mechanics: the attacker registers an application with a major identity provider (for example Microsoft Entra ID or Google Workspace) and requests broad delegated permissions. A lure (a shared-document email, or a message on a commercial messaging app) sends the victim to the provider's genuine login and consent page, with a valid domain and certificate. The victim authenticates normally, including MFA, then clicks Accept. The attacker exchanges the returned authorization code for an access token and a long-lived refresh token. Because the app calls APIs with a valid token, sign-in logs look clean, and a password change or MFA enrollment does not end access. Only revoking the grant or deleting the app does. Push Security describes the victim then seeing a generic 'File does not exist' error.

Permissions and abuse: Push Security describes file read/write, send-mail-as-user and mailbox-settings access. The latter allows a forwarding rule that copies all incoming mail to the attacker. Mail sent from the victim's account can be used for further phishing of colleagues. CSA cites Mail.ReadWrite, Files.ReadWrite.All and Calendars.ReadWrite as examples of broad scopes to review.

Related 2025-2026 activity reported by CSA (from its research note, not independently verified): the EvilTokens phishing-as-a-service kit (device code phishing, reported to have compromised 340+ Microsoft 365 organizations in five countries, surfacing mid-February 2026); ConsentFix, first documented by Push Security in December 2025, where v3 (May 2026) automates token exchange and reuses an existing Entra ID session so no password or MFA prompt occurs; TA4903 (Proofpoint-tracked, financially motivated) expanding into device code phishing in March 2026; the Salesloft/Gainsight OAuth token campaign; and a Vercel breach pivoting from a persisted Google Workspace OAuth grant to Context.ai. A secondary summary of an FBI IC3 PSA (dated September 1, 2026; identifier inconsistent in the secondary source) describes targeting of prominent individuals, their families and acquaintances via commercial messaging apps, with government, media and event-coordinator pretexts.

No attacker-owned IP addresses, domains or file hashes were published in any fetched source, so no BeaconBeagle C2 correlation was possible.

## MITRE ATT&CK

- T1566.002 Spearphishing Link
- T1566.003 Spearphishing via Service
- T1204.001 Malicious Link
- T1528 Steal Application Access Token
- T1671 Cloud Application Integration
- T1550.001 Application Access Token
- T1114.002 Remote Email Collection
- T1114.003 Email Forwarding Rule
- T1530 Data from Cloud Storage
- T1534 Internal Spearphishing

## Free detection (Splunk SPL, Microsoft KQL, Sigma)

**OAuth Consent Phishing — Broad-Scope Consent Grant (TL-2026-6177)** (`TL-2026-6177-SPL-001`)

Detects a user consent to an OAuth application in Entra ID whose payload contains broad delegated scopes (Mail.ReadWrite, Mail.Send, Files.ReadWrite.All, Calendars.ReadWrite, MailboxSettings.ReadWrite). This is the core consent-phishing event: the victim authenticates normally, clicks Accept, and the attacker app receives a refresh token that survives password resets and MFA.

- **Languages:** Splunk SPL, Microsoft KQL, Sigma
- **ATT&CK:** T1528, T1566.002, T1204.001
- **Data sources:** o365:management:activity (index=o365), Microsoft Purview Unified Audit Log, Entra ID AuditLogs
- **Severity:** high
- **Confidence:** high

### Splunk SPL

````spl
index=o365 sourcetype="o365:management:activity" Workload=AzureActiveDirectory Operation="Consent to application*" earliest=-24h
```TL-2026-6177 T1528 illicit consent grant: keep only user consent events whose payload carries broad delegated scopes```
| eval risky_scope=mvappend(if(match(_raw,"(?i)Mail\.ReadWrite"),"Mail.ReadWrite",null()), if(match(_raw,"(?i)Mail\.Send"),"Mail.Send",null()), if(match(_raw,"(?i)Files\.ReadWrite\.All"),"Files.ReadWrite.All",null()), if(match(_raw,"(?i)Calendars\.ReadWrite"),"Calendars.ReadWrite",null()), if(match(_raw,"(?i)MailboxSettings\.ReadWrite"),"MailboxSettings.ReadWrite",null()))
| where isnotnull(risky_scope)
| eval admin_consent=if(match(_raw,"(?i)IsAdminConsent.{1,40}True"),"true","false")
| stats earliest(_time) as first_seen latest(_time) as last_seen values(risky_scope) as risky_scopes values(admin_consent) as admin_consent values(ObjectId) as target_app by UserId, ClientIP
| convert ctime(first_seen) ctime(last_seen)
````

### Microsoft KQL

```kql
let riskyScopes = dynamic(["Mail.ReadWrite","Mail.ReadWrite.All","Mail.Send","Files.ReadWrite.All","Calendars.ReadWrite","MailboxSettings.ReadWrite"]);
AuditLogs
| where TimeGenerated > ago(24h)
| where OperationName =~ "Consent to application" and Result =~ "success"
| extend Actor = tostring(InitiatedBy.user.userPrincipalName), ActorIp = tostring(InitiatedBy.user.ipAddress), AppName = tostring(TargetResources[0].displayName), AppObjectId = tostring(TargetResources[0].id)
| mv-apply p = TargetResources[0].modifiedProperties on (
    where tostring(p.displayName) == "ConsentAction.Permissions"
    | extend Permissions = tostring(p.newValue))
| where Permissions has_any (riskyScopes)
| project TimeGenerated, Actor, ActorIp, AppName, AppObjectId, Permissions, CorrelationId
```

### Sigma

```yaml
title: OAuth Consent Phishing - User Consent to App With Broad Mail or File Scopes
id: 3b7e1c52-6177-4a01-9d1e-0a6c5f2b7a11
status: experimental
description: Detects a successful Entra ID consent grant whose permissions include broad delegated scopes (Mail.ReadWrite, Mail.Send, Files.ReadWrite.All, Calendars.ReadWrite, MailboxSettings.ReadWrite). Characteristic of illicit consent grants and consent phishing that yields a persistent refresh token without a password.
references:
  - https://pushsecurity.com/blog/consent-phishing-the-emerging-phishing-technique-that-can-bypass-2fa
  - https://labs.cloudsecurityalliance.org/research/csa-research-note-oauth-consent-phishing-ai-identity-2026052/
  - https://www.ic3.gov/PSA/2026/PSA260901
author: Threadlinqs Team
date: 2026-11-16
tags:
  - attack.credential-access
  - attack.t1528
  - attack.initial-access
  - attack.t1566.002
  - attack.execution
  - attack.t1204.001
logsource:
  product: azure
  service: auditlogs
detection:
  selection_op:
    properties.message: Consent to application
  selection_scope:
    properties.targetResources.modifiedProperties.newValue|contains:
      - 'Mail.ReadWrite'
      - 'Mail.Send'
      - 'Files.ReadWrite.All'
      - 'Calendars.ReadWrite'
      - 'MailboxSettings.ReadWrite'
  condition: selection_op and selection_scope
falsepositives:
  - Sanctioned productivity or CRM apps that legitimately request mail or file scopes
  - Admin-approved enterprise app onboarding
level: high

```

**Known false positives:**

- Sanctioned productivity, CRM or e-signature apps that request mail or file scopes
- Admin-approved enterprise app onboarding where an admin consents on behalf of the org

**Tuning notes:** Add an allow-list lookup of approved app display names or AppIds and exclude them. Raise severity when admin_consent=false and the app is unverified or newly registered. Extend the scope matches with organisation-specific sensitive scopes. Ensure the o365 audit feed includes Workload=AzureActiveDirectory.

### 8 more detection rules (Splunk SPL, Microsoft KQL, Sigma), unlock with TL Creds

- Device Code Flow Sign-In — EvilTokens Style Phishing (locked: unlock with TL Creds)
- First-Seen OAuth App Bulk Mailbox Access (locked: unlock with TL Creds)
- OAuth Token Redeemed From New IP After Consent (locked: unlock with TL Creds)
- New Service Principal, Credential or Permission Grant for OAuth App Persistence (locked: unlock with TL Creds)
- Scripted Bulk SharePoint/OneDrive Download With Application Token (locked: unlock with TL Creds)
- Claude Code Token File ([indicator]) Read by npm or Node Lifecycle Hook (locked: unlock with TL Creds)
- Mailbox Forwarding Rule Created After OAuth Compromise (locked: unlock with TL Creds)
- Internal Spearphishing Burst Through Graph/REST Token (locked: unlock with TL Creds)

Sign up free to unlock the detection rules: https://threadlinqs.com/?modal=signup

## Sample indicators (2 of 10)

Defanged, TLP:CLEAR.

- Domain: `microsoftonline[.]com`
- URL: `hxxps://microsoft[.]com/devicelogin`

Full set of 10 indicators unlocks with TL Creds: https://threadlinqs.com/tl-creds

## Sources

- [Consent phishing skips the password and walks in the front door (Expel) - body not retrievable](https://expel.com/blog/consent-phishing-skips-the-password-and-walks-in-the-front-door/)
- [Consent phishing: the emerging phishing technique that can bypass 2FA (Push Security)](https://pushsecurity.com/blog/consent-phishing-the-emerging-phishing-technique-that-can-bypass-2fa)
- [OAuth Consent Phishing: EvilTokens Bypasses MFA, Steals AI API Tokens (CSA research note)](https://labs.cloudsecurityalliance.org/research/csa-research-note-oauth-consent-phishing-ai-identity-2026052/)
- [OAuth Consent Phishing: The Attack a Password Change Does Not Fix (secondary summary of FBI PSA)](https://tuxxin.com/blog/oauth-consent-phishing-fbi-psa-2026)
- [FBI IC3 PSA on OAuth consent phishing (URL as cited by secondary source; not directly fetched)](https://www.ic3.gov/PSA/2026/PSA260901)
- [The app you approved in ten seconds (Abnormal AI; surfaced by search, not fetched)](https://abnormal.ai/blog/the-app-you-approved-in-ten-seconds)
- [Consent Phishing Account Takeover: Why a Password Reset Leaves the Attacker In (MojoAuth; surfaced by search, not fetched)](https://mojoauth.com/blog/consent-phishing-account-takeover-consumer-apps.md)

## Full data

The other 8 detection rules (Splunk SPL / Microsoft KQL / Sigma) unlock with TL Creds; a free account includes daily creds: https://threadlinqs.com/?modal=signup
Indicator values unlock with TL Creds, at a lower price on Red and Purple; AI agents can query rules and IOCs through the Threadlinqs MCP server (Purple): https://intel.threadlinqs.com/mcp

Canonical: https://intel.threadlinqs.com/threat/TL-2026-6177
