Adam Breuer (Harvard University), R. Eilat (Facebook Research), U. Weinsberg (Facebook Research)
Abstract: In 2019 alone, Facebook disabled over 6 billion fake user accounts. While early detection helps to reduce the harm that such accounts inflict, new fake accounts are notoriously difficult to detect via their pattern of social connections (i.e. 'friendships'), as their small number of connections are unlikely to reflect a significant structural difference from those of new real accounts. We present the SybilEdge algorithm, which determines whether a new user is a fake account ('sybil') by leveraging individual-level differences in how fake accounts interact with real users, and how real users react to fake accounts. Specifically, SybilEdge aggregates over each new user's (I) choices of friend request targets (i.e. users to whom she sends friend requests) and (II) these targets' respective responses (i.e. accept/reject). SybilEdge performs this aggregation giving more weight to a user's choices of targets to the extent that these targets are preferred by other fakes versus real users, and also to the extent that these targets respond differently to fakes versus real users. We show that SybilEdge rapidly detects new fake users at scale on the global Facebook network and outperforms existing state-of-the-art algorithms. To our knowledge, this is the first time a social network-based algorithm has been shown to achieve high performance (AUC>0.9) on new users who have only engaged in a small number of social interactions.