An Online Sockpuppet Detection Method Based on Subgraph Similarity Matching
Jingli Wang, Wei Zhou, Jiacheng Li, Yan Zhou, Jizhong Han, Songlin Hu · 2018
Multiple identity deception (sockpuppetry) has become an increasingly important issue in the social media environment, because they are engaged in undesired behaviors such as spreading rumors, posting hate speeches. The case of blocked users initiating new accounts, called sockpuppets, is widely known and past efforts, which have attempted to detect such users, have been primarily based on verbal features and non-verbal behaviors like posting habits. Although these methods have achieved a certain degree of success, it is easy for smart sockpuppets to forge verbal and behavioral features to circumvent management, making it difficult to guarantee the performance of these detection methods. In this paper, we observe that the social network plays a critical role in sockpuppets' influence, because recovering similar social structures can a sockpuppet guarantee the similar propagation impact. With this observation, a structure-based online method is proposed, which turns the sockpuppet detection problem into a subgraph similarity matching problem. The experiment on two real-world datasets of Sina Weibo demonstrates that our method obtains excellent detection performance, significantly outperforming previous methods.