Deepfake Candidate: An Applicant Used AI-Generated Video to Pass Remote Interviews

Background

A cybersecurity firm noticed subtle inconsistencies between a remote candidate's video interview appearance and their submitted government ID photo. The candidate had passed 4 rounds of technical interviews and was about to receive an offer for a senior security engineer role.

Investigation Methodology

1. Video Forensics: Recordings of the candidate's 4 video interviews were analyzed frame-by-frame for deepfake indicators: facial boundary artifacts, unnatural eye reflections, temporal inconsistencies, and head-pose alignment errors.
2. Identity Verification: The candidate's submitted identity documents were verified against government databases and compared with the video interview subject.
3. Background Correlation: The technical knowledge demonstrated in interviews was cross-referenced with the claimed work history of the person on the submitted resume.

Key Findings

• All 4 video interviews showed indicators of real-time deepfake rendering: micro-jitter at the facial boundary, inconsistent skin texture under varying lighting, and a 23ms latency pattern consistent with GPU-based face-swapping.
• The government ID belonged to a real person who was unaware their identity was being used.
• The actual person conducting the interviews was traced through IP metadata to a different country entirely.
• The scheme appeared designed to gain insider access to the cybersecurity firm's client infrastructure.

Outcome

The offer was not extended. The real identity document holder was notified. The incident was reported to the FBI as a potential nation-state espionage attempt. Total investigation time: 1 week.