Graph-based Sybil Detection in Social and Information Systems receives Best Paper Award

Yazan Boshmaf and his LERSSE research colleagues received a Best Paper Award at the  IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining for their paper, Graph-based Sybil detection in social and information systems. Defending against sybil attack is essential for today’s online sytstems. In this paper, Yazan and his co-authors present an analytical framework to evaluate graphic-based sybil Detection algorithms, a common method to defend against Sybil attack. 

Sybil Attacks

Yazan describes a Sybil attack in this way, “Any online user can join a system by providing an identity that is issued by either the system itself or by a trusted third-party. For example, Facebook users can join more than 9 million other websites and online services by simply authenticating with their existing user accounts Each user is intended to have a single identity and is expected to use this identity when interacting with other users in the system, but without tight offline-online identity binding these systems are vulnerable to the Sybil attack: the situation where an attacker forges many identities, each called a Sybil, and joins a target system for various adversarial objectives. For example, socialbots in online social networks control hijacked or adversary-owned user accounts in order to infiltrate these networks, steal private user data, spread misinformation, and distribute malware.” 

Summary of proposed framework
“Many methods to defend against Sybil attack model identity-based systems as graphs, where nodes represent identities and edges between nodes represent well-defined relationships (e.g., user profiles and mutual friendships in Facebook, respectively). These defence schemes utilize Graph-based Sybil Detection (GSD) algorithms to find identities that are likely to be Sybil based on their topological properties in the graph Even though many GSD algorithms exist, it is still unclear how these algorithms can be systematically evaluated, especially given that they make different assumptions about the used adversary and graph models. This paper presents a framework for evaluating GSD algorithms, along with an open-source implementation. The framework defines formal models to design, analyze, implement, and evaluate existing and new GSD algorithms under different adversary and graph models. The research team used this framework, along with a dataset of a real-world Sybil activity in Facebook.”

Read the paper:
Graph-based Sybil detection in social and information systems by Yazan Boshmaf, Matei Ripeanu, and Konstantin Beznoso, members of the LERSSE.

Find out more about the conference:

IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining The conference solicits experimental and theoretical works on social network analysis and mining along with their application to real life situations.