Structural-Spectral-Based Approach for Anomaly Detection in Social Networks

Abstract

Online social networks have become one of the most effective ways for connecting and communicating with people. These networks play a significant role in our business, social, and daily activities. In these networks, people follow a particular behavior that is not necessarily identical to the actual behavior in their real life. Our goal in this work is to investigate and explore the interactions among people in the Facebook network as our targeted social network. Our investigation aims to detect the potential anomalous behaviors within the interactions among people. To this end, we involve the structural and spectral features of the network in proposing a new approach for anomaly detection. Besides, our approach is supported by concepts that are inspired from sociological theories. The data of this article was extracted from the Facebook network using Facebook Graph API. In the experimental results, the proposed approach reflected an efficient performance in terms of detecting potential anomalies and computational complexity compared to other approaches in the literature.