Trust in Social Network-based Sybil Defenses

• We first measure the mixing time of different social graphs, and observe a relationship between the mixing time and the level of knowledge (trust) in the underlying graph. • Then, we learn the impact of the different proposed designs on the mixing time. We show that parameters associated with the different mixing models for characterizing trust control the mixing time. • Finally, we learn the cost of SybilLimit to accept all non-Sybil nodes in some social graphs, under varying parameters for the different designs. While these proposed designs characterize trust, we show that trust—once it is incorporated into the Sybil defense—comes at some cost. The datasets (n/deg/μ) are below.