How a ClickFix Scam Could Turn an AI Search into Microsoft 365 Business Email Compromise

· Original article ↗

Summary

Huntress illustrates a fictional attack chain in which a malicious custom GPT leads an employee to run a ClickFix command, enabling session-cookie theft and invoice fraud, and describes endpoint and identity clues defenders could watch for.

Key points

  • In Huntress’s fictional scenario, an employee reaches an attacker-built custom GPT through a sponsored search result, then follows its instructions to run a command in Terminal.
  • The ClickFix command compromises the employee’s Mac; the attacker retrieves browser cookies and reuses the Microsoft 365 session cookie to access Outlook without the password or a new MFA prompt.
  • The attacker creates a hidden inbox rule to archive invoice-related messages, impersonates the employee in a payment-change request, and deletes evidence; the scenario unfolds in about 24 hours.
  • Potential endpoint clues include a Mac shell running curl and an attacker IP address encoded in the command.
  • Potential identity clues include two browsers or operating systems associated with one session, unexpected VPN traffic, and an inconspicuous inbox rule.
  • Huntress’s Tragic Quadrant compares tactics by prevalence and potential impact, emphasizing how lower-impact techniques can become more damaging when attackers chain them together.

Article Details

Topic
Illustrative ClickFix-to-business-email-compromise attack chain involving session-cookie theft and hidden inbox rules

MITRE ATT&CK

People

Vendors

Products

Related Articles