Invitation or Bait? Detecting Malicious URLs in Facebook Events

With 2.2 billion monthly active users, Facebook is the most popular Online Social Network. Given its huge popularity and diverse features such as pages, events, groups etc., it is potentially the most attractive platform for cybercriminals to launch various attacks. In this paper, we study the role of Facebook Events in disseminating malicious URLs in the network. Here, we focus our analysis on Facebook Events which are created by Facebook Pages. The existing services like Web of Trust (WOT) and other blacklists follow crowdsourcing models. Thus, malicious URLs can only be detected once they are widespread on the network and has done significant damage. Therefore, we train a supervised machine learning model on our labeled dataset to create an efficient classifier for automatic detection of malicious Facebook events, independent of blacklists and third-party reputation services. Our model is able to classify malicious events with 75% accuracy using Support Vector Machine. To the best of our knowledge, this is the first paper to study the presence of malicious URLs on Facebook Events.