Detecting Memory Editing Cheats by Validating Host Memory Integrity from GPU

As the financial size and impact of the video game industry grows, so does the impact of cheating in video games. For example, cheating in online multiplayer games could affect the fairness of their championships, which award tens of thousands of USD. A unique challenge in cheat detection for video games is that the owner of the system (on which the games run) are the adversaries, thus they have full control over the entire system. To achieve cheat detection against such adversaries, we propose a novel cheat detection system based on host memory monitoring from GPU. It monitors the value of protected objects and compares them with pre-loaded metadata, and detects cheats when they do not agree with each other. By monitoring the host memory by a GPU program that cannot be intercepted once invoked, it achieves resiliency against a cheater who has a kernellevel privilege of the system. Our evaluation shows that it can successfully find two types of cheats and that it is robust to a cheater's attempt to overwrite the metadata.