Improving Social Network-Based Sybil Defenses by Rewiring and Augmenting Social Graphs

摘要

Recent solutions to defend against the Sybil attack, in which a node generates multiple identities, using social networks. In these solutions, social networks are assumed to be fast mixing, and Sybil nodes--which disrupt the fast mixing property of social networks--are detected. Little is known about the cause of the mixing quality in social graphs, and how to improve it in slow mixing ones. In this work we relate the mixing time of social graphs to graph degeneracy, which captures cohesiveness of the graph. We experimentally show that fast-mixing graphs tend to have a larger single core whereas slow-mixing graphs tend to have smaller multiple cores. We then propose several heuristics to improve the mixing of slow-mixing graphs using their topological structures by augmenting them. We show that our heuristics greatly improve Sybil defenses.