Detecting Sybil Nodes in Anonymous Communication Systems.

Abstract: As one of the most fundamental problems in open-membership systems, Sybil attack has attracted wide attentions from both industrial and academic fields. Many anonymous communication and censorship-resistant systems have incorporated Sybil defense in their designs based on social networks. This paper proposed two-class undirected mixed membership stochastic blockmodels to discover adversary's Sybil identities in the network. Different from existing algorithms, our model discriminates Sybil nodes by simulating the generative process of social networks. And we gave a matrix B to describe the interaction probability of Sybil and honest nodes, which can be used to observe contaminated situation of the network. The effectiveness of the algorithm is validated by experiments over both simulated and real-world social networks. The method can be also used to further improve the accessibility of anonymous and anti-censorship systems. [Copyright &y& Elsevier]