Cheat Detection in On-Line Multi-Player Games using Graphics Processing Units

Abstract

On-line multi-player games have experienced an impressive growth over the last decade. Despite this success, game providers struggle to keep up with the many different types of cheating occurring in these games. Due to the computational demand of a large scale multi-player on-line game, server resources are becoming scarce. This makes the task of implementing cheat detection mechanisms difficult, because of the lack of computational resources. Advances within the field of General Purpose computing on Graphic Processing Units (GPGPU), have given developers easier access to the computational power of the GPU. In this thesis, we investigate what possible benefits there are of implementing a GPGPU cheat detection mechanism. We have developed a framework for a game simulator that includes a simple customizable physical engine and a cheat detection mechanism. We have created both a CPU and a GPGPU version of the cheat detection mechanism we have constructed. The GPGPU implementation runs on NVIDIA GPUs using the Compute Unified Device Architecture (CUDA) framework. We have also constructed a simple user interface to provide a graphical representation of the game simulator. The results we have obtained from our research indicate that offloading cheat detection mechanisms to the GPU, increases the speed of the mechanism. We also discover that in addition to being faster, the GPU mechanism allows the Central Processing Unit (CPU) to perform other game relevant tasks while the mechanism is executing. Overall, our research shows that game providers can benefit from offloading certain parts of their server side processing to the GPU.