Modern video games are becoming increasingly complex, particularly in the realm of multiplayer first-person shooter (FPS) games, which highlights the necessity for effective anti-cheat mechanisms. This paper focuses on detecting players using Aimbot cheats, identifying specific features from both time-series data (such as aim velocity, number of shots, distance to target) and behavioral data (including utility usage and player movement patterns).
We developed a server-side aimbot detection classifier named 'YAACS', which consists of a parser, a deep learning model, and intermediary connection utilities for integration with the game server. By analyzing sequences of 128 ticks, YAACS achieved a classification accuracy of 88.6% with a false positive rate of just 0.97%. In contrast, the Decision Tree model reached a higher accuracy of 96.2% but had a false positive rate of 2.68%, making it 2.76 times worse than the best LSTM configuration. These results underscore the importance of incorporating temporal context through sequence modeling to minimize false accusations in FPS cheat detection.
Blogger's Review: This paper addresses the Aimbot detection challenge in FPS games using deep learning, highlighting the significance of time-series data, which offers new insights and methodologies for future anti-cheat systems.