FakeBuster: A Robust Fake Account Detection by Activity Analysis

Abstract

Online social network is playing an important role in modern society. It not only significantly changed the way people communicate, but also provided a new way of potential attack, such as identity theft, false information, etc. Recently, fake accounts in online social networks (OSNs) have made several problems throughout online social network. With the massive creation of fake account, OSNs providers are suffered by overflooded advertisement, false information spread, etc. Traditional method such as cannot distinguish between real and fake account efficiently. Some previous work has studied the structure and composition of fake account that distribute spam messages, however the fast improvement of fake account creation made these previous works outdated or getting ineffective. We then target on new model of fake account that may not only automatically post or comment, but also spam with advertisement or spread false information. In this research, we proposed an innovative method to detect fake account in OSNs. This method based on user activity pattern through Facebook activities, and we leverage machine learning to predict if an account is controlled by fake user. We also provide an analysis of those account activity from the prediction result. This research may bring up to a whole new level in fake accounts detection