Research On Sybil Attack Detection Algorithm Based On Random Walks Betweenness In Social Networks

With the rapid development of social networks, more and more users communicateand share information via social networks, but due to the openness of social networks,social network users are more vulnerable to security threats, especially the upward trendof Sybil attack, a malicious user creates a large number malicious identities in socialnetworks in search of benefits, spreading malicious information to honest users or gainmore influence in the system, seriously threating the social networks security. In this paper,based on a comprehensive analysis of research status at home and abroad, an in-depthresearch is conducted on how to solve the problem of Sybil attack detection on large scalesocial networks.Firstly, the existing Sybil attack detection algorithm has relatively strict assumptionsand cannot be applied effective in the large-scale social networks due to their algorithmsare computationally intensive, an analysis of the characteristics of Sybil attack model iscarried out, the betweenness of an edge is the number of paths among peers passing theedge, and the edge betweenness of attack edge is significantly higher than that of honestedge, a kind of edge betweenness model is proposed which is feasible for large scalesocial networks and more close to the real information propagation in social networks,named c-path edge betweenness model, and an algorithm is proposed to calculate theedge betweenness by liming the length of path and choosing reasonable strategy.Secondly, in view of the existing attack detection algorithm has no effective schemeto detect malicious user group, this article proposes a clustering-based approach todetecting the Sybil community. The algorithm combines the edge betweenness and theedge clustering coefficient as characteristics for edge clustering, and then determines thereal edges cluster and Sybil edges cluster depend on the number of honest user from theseed data set. Then a method is proposed to detect malicious groups using an improvedlabel propagation algorithm.Finally, we compare the experimental evaluations and analysis of the algorithmsproposed in this paper with the traditional methods on data sets with different scale.