In this paper, we propose an approach to detect cheating with game bots in Massively Multiplayer Online Role-playing Games (MMORPGs). Cheating with game bots which can auto-play online games without human involvement is a big threat to the industry of MMORPGs. The proposed approach can detect game bots by analysis of encrypted game traffic. In the proposed detection, hidden Markov models (HMMs), known for the power in temporal pattern recognition, are trained to model game bots' and human players' gaming behaviors. A detection decision is made by testing a game trace of interest against trained HMMs. We evaluate the proposed detection approach with game traces collected from the Internet. Our experiment results show that the proposed detection can detect game bots accurately with only a small number of training traces.
展开▼
