IDENTIFICATION AND ANALYSIS OF FAKE IDENTITIES ON OSN

Abstract

Online social networking site suffer from the usage of fake accounts that leads to fake product reviews, advertise, malware and spam. Existing system focus on using the social graph approach to detect fakes. However, our project shows that fake user could be friend of a large number of genuine users, invalidating the assumption of social graph based detection. In this project, we represent VoteTrust, a reliable system that further protect user level activities. VoteTrust models the friend request scenario among users as a directed, signed graph, and use two key mechanisms to find fake user over the system : a votingbased fake detection to find users that other users vote to reject, such that fake user community detection to find other colluding fake around identified Sybil’s. Through evaluating on social network, we show that VoteTrust is able to prevent user from generating many irrelevant friend requests.