Hybrid graph-based Sybil detection with user behavior patterns
Xiang Li, Qixiao Lin, Jian Mao · Procedia Computer Science · 2021
Online social networks (OSNs) are known to be vulnerable to Sybil Attack, where attackers leverage the openness to create multiple fake identities for launching many malicious activities. In this paper, we define a weighted-strong-social (WSS) graph that integrates the OSN structure and user behavior patterns and propose a novel hybrid graph-based sybil detection approach. The hybrid approach estimates the trustworthiness of users and user pairs based on user behaviors that can be obtained locally and add them to the OSN structure to construct a WSS graph for sybil detection. The evaluation results show that the AUC of the hybrid approach is 0.954, which is significantly higher than that of previous sybil detection methods.