A student from the Norwegian University of Science and Technology created a system to identify CS2 players through their behavior using match demo analysis. The core idea aims to ban the actual player rather than a specific account, preventing cheaters from returning with new profiles.

The system utilizes two independent signals from existing match demos: mouse movements and keyboard presses. Mouse movement analysis correctly identified players in every case out of a test sample exceeding 1,000 players.

Keyboard analysis demonstrated a 98% accuracy rate, while combining both signals achieved a 100% match rate during tests. Furthermore, the method automatically discovered unknown smurf accounts by correlating behavioral profiles with Steam friend lists and activity.

An optimized version can run on a single 20GB GPU A100 segment to handle thousands of new demos generated hourly. Limitations involve a small tested sample of duplicate accounts and the requirement for new accounts to play several matches first. The developer urges Valve and matchmaking platforms to test the system on a real player audience.