Detecting Suspicious Following Behavior In Multimillion-Node Social Networks

In a multimillion-node network of who-follows-whom like Twitter, since a high count of followers leads to higher profits, users have the incentive to boost their in-degree. Can we spot the suspicious following behavior, which may indicate zombie followers and suspicious followees? To answer the above question, we propose CATCHSYNC, which exploits two tell-tale signs of the suspicious behavior: (a) synchronized behavior: the zombie followers have extremely similar following behavior pattern, because, say, they are generated by a script; and (b) abnormal behavior: their behavior pattern is very different from the majority. Our CATCHSYNC introduces novel measures to quantify both concepts and catches the suspicious behavior. Moreover, we show it is effective in a real-world social network.