Trust-based supernode selection in peer-to-peer systems

The P2P applications employing supernode architecture become more and more popular, while the problem about supernode selection is still open. When selecting supernodes, current researches and application fields mainly focus on the capability criterion such as software and hardware. So this provides malicious node with a good chance to exert malicious behaviors because they can seriously damage the system by upgrading their capability to become supernodes. This paper presents a new trust-based supernode selection method that is complementary to the only-capability selection. We use the concept of proxy trust to identify the behaviors of supernodes. With this metric a peer can select its own trusted supernodes and isolate malicious supernodes in the system. Furthermore in order to reduce network cost, we do not adopt basic flooding mechanism, but a limited one to distribute the trusted supernodes list requests. Simulations show our method can isolate malicious supernodes within a rather short period and greatly improve the probability of successful services. At the same time network overhead is small in our method.