BEAT: Behavior Evaluation and Anomaly Tracking, Game Bot Detection Framework in RPG Games

Abstract

In the realm of gaming, game bots represent a serious threat to the game balance, resulting in decreased enjoyment for legitimate players. For developers, the infiltration of bot players can lead to substantial loss of normal players, ultimately shortening the game’s lifespan and reducing revenue. The proliferation of game bots has significantly impacted the gaming industry’s ecosystem. Accurately identifying game bots and combating black market production is critical for ensuring secure game operations. However, the number and variety of RPG game bots has been on the rise in recent years, posing a significant challenge for security operations and making bot detection increasingly difficult. To this end, we propose BEAT: behavior evaluation and anomaly tracking, game bot detection framework in RPG games. It combines the game settlement data, click trajectory and movement trajectory data of players in the game behavior trajectory, generates multi-view data of players from multiple angles, and uses supervised and semi-supervised models to detect game bots. And the corresponding explanation of the model is given. BEAT has been deployed and applied in a number of domestic and foreign RPG games, greatly improved the detection efficiency, also achieved a very significant effect. Finally, our framework realized automatic model update iteration, which can not only detect known game bot types, but also automatically identify some new game bot types.