Modelling reputation-based resource pooling in P2P systems

P2P systems have a very loose social structure and lack a central authority that enforces cooperative behavior. If peers choose not to contribute but to free ride, detecting and punishing misbehaviour is hard. Several solutions have been proposed that make anti-social behavior less profitable, all with different applications in mind. The two basic categories are payment schemes and reputation mechanisms.In this paper, we propose a new model for P2P systems and use it to analyse the effects of different reputation mechanisms on performance of the application. We identify basic classes of peers, reputation functions, resource distribution strategies, and adversary types, and present simultation results for a few systems composed of these building blocks. Finally, we expose an antagonism between the goal of establishing a distributed data structure with mostly noise-free reputation information and the goal of punishing free riders that all methods will have to face.