Understanding Sybil Groups in the Wild

Abstract

Sybil attacks are one kind of well-known and powerful attacks against online social networks (OSNs). In a sybil attack, a malicious attacker generates a sybil group consisting of multiple sybil users, and controls them to attack the system. However, data confidentiality policies of major social network providers have severely limited researchers’ access to large-scale datasets of sybil groups. A deep understanding of sybil groups can provide important insights into the characteristics of malicious behavior, as well as numerous practical implications on the design of security mechanisms. In this paper, we present an initial study to measure sybil groups in a large-scale OSN, Renren. We analyze sybil groups at different levels, including individual information, social relationships, and malicious activities. Our main observations are: 1) user information in sybil groups is usually incomplete and in poor quality; 2) sybil groups have special evolution patterns in connectivity structure, including bursty actions to add nodes, and a monotonous merging pattern that lacks non-singleton mergings; 3) several sybil groups have strong relationships with each other and compose sybil communities, and these communities cover a large number of users and pose great potential threats; 4) some sybil users are not banned until a long time after registration in some sybil groups. The characteristics of sybil groups can be leveraged to improve the security mechanisms in OSNs to defend against sybil attacks. Specifically, we suggest that OSNs should 1) check information completeness and quality, 2) learn from dynamics of community connectivity structure to detect sybil groups, 3) monitor sybil communities and inspect them carefully to prevent collusion, and 4) inspect sybil groups that behave normally even for a long time to prevent potential malicious behaviors.