Bot detection in rhythm games

Abstract

As the online game industry expands, detecting and preventing cheating in games is an increasingly important research topic. Some forms of cheating, such as the use of game bots (auto-playing game clients), are particularly challenging to identify because game bots do not violate any of the game rules; rather, they simply mimic human behavior to play the game without human intervention. The use of bots introduces fairness issues to online games, and therefore robust schemes for detecting game bots are strongly demanded.In this paper, we tackle with bots in rhythm games, which feature gameplay that incorporates eye and body coordination with music, usually a popular song. Bot detection in rhythm games is especially challenging compared with in other game genres because little information is available to distinguish the responses made by a human player from a bot. Based on the long-memoryness of the time series formed by human players' response errors to stimuli, we propose a scheme to detect the presence of human coordination mechanisms during gameplay. Based on a set of traces collected from human players and real-life game bots, we show that our scheme can accurately detect the use of game bots despite of game difficulty levels.