Social Turing Tests: Crowdsourcing Sybil Detection
Social Turing Tests: Crowdsourcing Sybil Detection Gang Wang, Manish Mohanlal, Christo Wilson, Xiao Wang‡, Miriam Metzger†, Haitao Zheng and Ben Y. Zhao Department of Computer Science, U. C. Santa Barbara, CA USA †Department of Communications, U. C. Santa Barbara, CA USA ‡Renren Inc., Beijing, China Abstract The research community has produced a substantial number of techniques for automated detection of Sybils [4, As popular tools for spreading spam and malware, Sybils 32, 33]. However, with the exception of SybilRank [3], few (or fake accounts) pose a serious threat to online communi- have been successfully deployed. The majority of these ties such as Online Social Networks (OSNs). Today, sophis- techniques rely on the assumption that Sybil accounts have ticated attackers are creating realistic Sybils that effectively difficulty friending legitimate users, and thus tend to form befriend legitimate users, rendering most automated Sybil their own communities, making them visible to community detection techniques ineffective. In this paper, we explore detection techniques applied to the social graph [29]. the feasibility of a crowdsourced Sybil detection system for Unfortunately, the success of these detection schemes is OSNs. We conduct a large user study on the ability of hu- likely to decrease over time as Sybils adopt more sophis- mans to detect today’s Sybil accounts, using a large cor- ticated strategies to ensnare legitimate users. First, early pus of ground-truth Sybil accounts from the Facebook and user studies on OSNs such as Facebook show that users are Renren networks. We analyze detection accuracy by both often careless about who they accept friendship requests “experts” and “turkers” under a variety of conditions, and from [2].
[Show full text]