Nowadays, Online Social Networks (OSNs) has become one of the most common ways among people to facilitate communication. This has made it a target for attackers to steal information from influential users and has brought new forms of customized attacks for OSNs. Attackers take advantage of the user's trustworthiness when using OSN. This exploitation leads to attacks with a combination of both classical and modern threats. Specifically, colluding attackers have been taken advantage of many OSNs by creating fake profiles of friends of the target in the same OSN or others. Colluders impersonate their victims and ask friend requests to the target in the aim to infiltrate her private circle to steal information. These types of attacks are difficult to detect in OSNs because multiple malicious users may have a similar purpose to gain information from their targeted user. The purpose of this paper is to overcome this type of attack by addressing the problem of matching user profiles across multiple OSNs. Then, we will extract both features and text from a user's profile and build a classifier based on supervised learning techniques. Simulation and experimental results are provided to validate the accuracy of our findings.
展开▼
