Attribute Credibility Based Sybil Goup Detection in Online Social Networks
Yechao Xia, Pan Li, Liang Shi, Futai Zou · 2016
Recent years, online social networks (OSNs) are facing increasingly serious Sybil attack, which results in unexpected terrible impacts on social network services with illegal behaviors as manipulating online votes, publishing large amount of annoying advertises, etc. To detect Sybil account among ordinary users, several attributes and behavioral characteristics based detection algorithms were proposed, most of them claimed to have desirable results in corresponding experiments. Among these well designed algorithms, VoteTrust deserves special attention, which successfully reduced high false detection rate due to Sybil users' invasion to real users' social community. However, this algorithm cannot deal with real users' relatively high false follow rate in follow-forwarding social networks. In this paper, an attribute credibility based Sybil group detection approach is proposed to tackle the drawback of VoteTrust mentioned earlier. It applies Euclidean distance between user's attribute and the center of Sybil attribute as a measure to calculate the credibility, then utilize the value as a key parameter to obtain user's trueness, finally Sybil group can be classified by trueness. Extensive experiments have been conducted based on Sina Weibo dataset. The results show that the proposed approach based on the attribute and behavior characteristics has higher detection accuracy than state-of-the-art related algorithms.