Sherlock is an AI-driven interview integrity platform designed to monitor, analyze, and secure online interviews. By evaluating video, audio, and behavioral feeds, Sherlock helps recruiting and hiring teams identify potential cheating, deepfakes, or unauthorized assistance, ensuring a fair evaluation process for every candidate.
During a scheduled interview, Sherlock joins the meeting as a silent participant or runs in desktop mode to observe and protect the integrity of the session.
1. Multi-Signal Detection
While the interview is active, Sherlock’s AI engine runs in the background, analyzing the live feed across multiple vectors. The detection report is generated after the interview.
🔹 Detect mode/Real‑Time Deepfake + Detect
Only Deepfake Detection operates in real time.
Real‑Time Deepfake Detection: Scans the stream to detect synthetic video, digital camera overlays, or face‑swapping software.
Visual Behavioral Analysis: Monitors visible gaze direction, off‑screen attention, screen‑share context, answer timing, content appearance patterns, and audio cues where observable.
Audio and Voice Tracking: Flags secondary background voices, whispering, or external human assistance.
🔹 Desktop Mode
Desktop Mode focuses on system‑level activity detection. It monitors local system behavior to identify suspicious application usage or unauthorized background processes during the interview.
2. The Integrity Score & Flagging System
Once the interview concludes, Sherlock compiles the raw analysis into a detailed integrity report. The report highlights trust signals, anomalies, and overall risk level for each session.
Scoring Scale (0 to 10): Every completed interview receives an Integrity Score. A higher score indicates a clean, flag‑free session.
Report Summary: Each report includes visual indicators of risk level, detected anomalies, and recommendations for peer validation.
