Using multiple verifiers to detect sybils in a social network graph

Abstract

Detecting Sybil identities is important to operate a distributed system without losing its openness property. Recently, OSN(Online Social Network) based Sybil detection methods are proposed and an individual node can determine whether other nodes are Sybil or not. However, since the probabilistic properties of the previous methods, single verifier based Sybil detection may suffer from the wide variance in the performance of detecting Sybil nodes. In this paper, multi-verifier based Sybil detection method is proposed to mitigate the variance. The proposed method selects honest verifiers from a social network graph where honest and Sybil nodes are mixed. Then, the method determines whether a node is Sybil or not by comparing the likelihood of acceptance of the node and a given threshold. Through the extensive evaluation with the real-world social network sample graph, the proposed multi-verifier based Sybil detection outperforms the single verifier based Sybil detection in both aspects of accepting honest nodes and suppressing Sybil nodes. © 2014 Springer-Verlag.