Automatic ICA detection in online social networks with PageRank

Nowadays, online social networks have become an essential part of humans. However, there are some dark side to this widespread use of online social networks. One of them is the fact that many attackers have succeeded to clone celebrities' profiles and have attracted hundreds or thousands of followers. This type of forging has caused many problems for famous people. This phenomenon is commonly known as Identity Cloning Attack which is abbreviated to ICA in the literature. ICA occurs when a malicious user selects one of the famous users of a social network as his victim. The attacker then creates a user account similar to the victim's profile and embarks on various malicious social activities. In this paper, we have proposed an automatic method to identify cloned profiles. This method consists of three main steps and is implemented on Hadoop framework using the MapReduce programming model. In the first step, we count the number of followers of each user and store it as an attribute for their profile. In the second step, the network users are clustered based on their profile attributes and their number of followers. Subsequently, we move all the profiles within the same cluster as the victim's profile to the next step and consider them as suspicious profiles. The victim's profile is a profile of a celebrity, where the proposed method is conducted to verify its authenticity. In the third step, we eventually select the profile with the highest rank as the valid profile. This method of ranking the profiles is based on the outcome of PageRank algorithm in the first step. This method is easily applicable and does not require any additional information for identifying the original user account. Furthermore, this method employs a distributed processing framework, limits the search space, and decreases the required computation by clustering the profiles. We have applied the suggested method to a dataset that we collected from Instagram. Our findings were quite promising, and in some situations, we were able to identify all the cloned profiles with a 100% accuracy. The results are comparable to the best ones in this area of study.