Trust management for cheating detection in distributed massively multiplayer online games

In this paper we examine the use of trust metrics for punishing dishonest (cheating) and malicious (griefing) behavior in peer-to-peer massively multiplayer online games (MMOG’s). In particular, we first replicate the work of Goodman and Verbrugge and then propose a metric of our own. Our approach uses a reinforcement learning trust metric that can more rapidly and accurately detect dishonest and malicious behavior. We use a genetic algorithm which keeps in mind the goals of the system to select the parameters for our trust metric. Our metric produces a system which removes dishonest and malicious players from the system much faster than the original while nonetheless retaining a low false positive rate. The metric also produces a surprising result in which cheating and griefing are treated equally, despite being unequal offenses.