Cheat detection for games using runtime monitoring

Abstract

Since cheating in games has becomes more prevalent, companies should aim to detect cheaters so that the experience of honest players is not affected. Failure to punish these cheaters can potentially result in a fi nancial loss for the developers. This project aims to investigate a runtime monitoring approach to cheat detection in shooter games where the cheat detector is written separate from the game code. The approach aims to increase code maintainability due the separation of the cheat detector from the game code. This approach allows the cheat detector to be reusable while also reducing the chance of introducing bugs into the bot detector when re-factoring the game code. Due to this separation, components are weaved together and run concurrently at runtime. The main disadvantage of this approach is the additional computing time required (overheads) which can make the approach impractical. This approach detects cheaters depending on the player's behaviour. Throughout the execution of the game the player's behaviour is monitored and once the user stops playing a cheating score is obtained along with a decision on whether the player is deemed to be cheating or not. Two use cases have been designed for this project. The first use case is a proof of concept which aims to verify whether the proposed approach can successfully extract game events while monitoring the game's execution. Subsequently, a larger use case has been designed which is comprised of two similar games thus simulating the typical evolution of a game. A cheat detector has been written and adapted to work on both games. This use case aims to show that the separation-of-concerns approach aids reusability in systems notwithstanding certain changes in the game code. The main objective of the research is to verify the advantages of the separation-of-concerns approach (embodied in the runtime monitoring approach) on the modules produced by de ning certain software metrics including cyclomatic complexity and the maintainability index. Additionally, the overheads produced and the detection rate of the cheat detector have been evaluated but these do not effect the viability of the approach since they are dependant on implementation decisions. Additional evaluations have been performed on the resuability between modules and the advantage of language independence between modules. The evaluation showed that the code produced is more maintainable and less complex. Additionally, the accuracy of the bot detector is of 85% while the overheads produced did not negatively impact the experience of the user.