Anti-Cheat System Deep Dive
How we detect and prevent assessment fraud
TalentScreen's anti-cheat system uses behavioral analysis, device fingerprinting, tab monitoring, and AI detection to identify suspicious activity. Multiple signals are combined to generate confidence scores.
Detection Signals
- Tab switching - Leaving assessment window
- Copy/paste events - Pasting answers from external sources
- Typing patterns - Unnatural speed or rhythm
- Answer timing - Too fast or suspiciously consistent
- IP address changes - Mid-assessment location switches
- Device switching - Starting on one device, finishing on another
- Similar answers - Matching other candidates (collusion)
- AI content detection - LLM-generated responses
- Mouse movement - Bots vs. human patterns
Confidence Levels
Low confidence (1-2 signals) generates a warning. Medium confidence (3-4 signals) flags the session for review. High confidence (5+ signals) marks the assessment as compromised and recommends disqualification.
False Positive Prevention
Single events rarely trigger flags. The system looks for patterns and corroborating evidence. For example, one tab switch may be accidental, but 15 tab switches combined with fast answers and copy events suggests cheating.
Always review evidence manually before disqualifying candidates. False positives, while rare, can occur with legitimate browser behavior.
Was this article helpful?