A Graph Based Approach for Fake User Accounts Detection in Social Networks

Nowadays Online Social Networks regularly face a severe threat by the continuous creation and widespread activity of fake user profiles, which can undermine the quality of social network service by introducing spam and malicious techniques to promote low-rated products and services, spread fake news or even steal one’s personal information and money. Recently, there has been much excitement among researchers over exploiting social network structure to detect such fake profiles. However, they generally rely on the assumption that fakes form a very dense community, which may not hold in real OSNs. In this research, a graph- based algorithm for fake user detection is presented. The algorithm uses the techniques of social network topology construction and communities’ detection to evaluate if an account can be considered as fake. Moreover, Machine Learning classification algorithm is used to obtain the probability of a particular account is fake. Using detailed evaluation of real social network VK which is highly prevalent in Russia and CIS countries, we show that algorithm is able to detect fake user accounts with quite a high accuracy and precision.