A New Automatic Method for ICA Detection in Online Social Networks
Abstract
Online information of social network users, which is shared inside these networks, has started to be interesting for malicious users. One of the threats to users` personal information is called Identity Clone Attack (ICA). In this type of attacks, the malicious user targets an online social network user as a victim and creates a fake but similar profile to his profile. In ICA, most of the time, famous and beloved users of the society like celebrities and politicians are being preyed on. This paper presents a new automated method for detecting fake profiles cloned from celebrities’ profiles. In this method, we first cluster the profiles. Subsequently, the profiles which are in the same cluster as the victim’s profile with a similarity to the victim`s profile over a predefined threshold are moved to the next phase as suspicious profiles. Then a parameter named “Celebrities’ Network” is extracted for the suspected and the victim’s profiles. “Celebrities’ Network” for a user is a sub graph of the network, whose nodes are popular users following that user. The profile for which this parameter is the highest is considered as the real profile. This is an easily applicable method as it does not require a human agent and extra information. Besides, initial clustering helps the system to save time in searching. The suggested method was applied to face book and Instagram datasets and approximately 100% of the cloned profiles were detected. These findings show that this method is quite promising and the results are comparable to the best studies conducted on ICA.
Keywords
Online Social Network, Fake Profile, ICA, Threat, Popular Users
