Виявлення аномальної поведінки гравців засобами комп’ютерного зору

Abstract

This study examines the application of computer vision methods for detecting anomalous player behaviour in first-person shooter games. Traditional signature-based anti-cheat systems demonstrate limited effectiveness against modern cheat software capable of bypassing file-level detection. The proposed approach employs real-time object detection algorithms, particularly YOLO, to identify game entities and analyse cursor movement patterns. Deviations from statistically typical human motor behaviour serve as indicators of automated assistance software. The method offers a non-invasive alternative to kernel-level anti-cheat solutions while maintaining detection capability for both aimbot and wallhack-type violations.