On Sybil Classification in Online Social Networks Using Only Structural Features

Abstract

Sybil attack is a problem that seriously affects Online Social Networks (OSNs). These attacks are made possible by the openness of OSN platforms that allows an attacker to create multiple fake accounts, called Sybils, which are then used to compromise the underlining trust pinnings of the OSN. Early Sybil account detection mechanisms involved classification of users into benign and malicious based on various attributes collected from the user profiles. One challenge affecting these classification methods is that user attributes can often be in-complete or inaccurate. In addition, these classification methods can be evaded by sophisticated attackers. More importantly, user profiles can often reveal sensitive user information that can potentially be misused causing privacy violation. In this work, we propose a Sybil detection method that is based on the classification of users into malicious and benign based on the inherent topology or structure of the underlining OSN graph. We propose a new set of structural features for a graph. Using this new feature set, we perform several experiments on both synthetic as well as real-world OSN data. Our results show that the proposed detection method is very effective in correctly classifying Sybil accounts without running the risk of being evaded by a sophisticated attacker and without compromising privacy of users.