Generic anomalous vertices detection utilizing a link prediction algorithm

• Published in: Social network analysis and mining Vol. 8; no. 1; p. 27
• Main Authors: Kagan, Dima, Elovichi, Yuval, Fire, Michael
• Format: Journal Article
• Language: English
• Published: Vienna Springer Vienna 01.12.2018; Springer Nature B.V
• Subjects: Algorithms; Apexes; Applications of Graph Theory and Complex Networks; Co authorship; Computer networks; Computer Science; Data Mining and Knowledge Discovery; Datasets; Economics; False positive results; Game Theory; Graph theory; Graphs; Humanities; Law; Machine learning; Methodology of the Social Sciences; Methods; Original Article; Social and Behav. Sciences; Social networks; Statistics for Social Sciences; Topology; Anomaly Detection; Transit Friendly; Classical Link Prediction; Fake Profiles; Anomalous Vertex
• ISSN: 1869-5450; 1869-5469
• DOI: 10.1007/s13278-018-0503-4

In the past decade, graph-based structures have penetrated nearly every aspect of our lives. The detection of anomalies in these networks has become increasingly important, such as in exposing infected endpoints in computer networks or identifying socialbots. In this study, we present a novel unsupervised two-layered meta-classifier that can detect irregular vertices in complex networks solely by utilizing topology-based features. Following the reasoning that a vertex with many improbable links has a higher likelihood of being anomalous, we applied our method on 10 networks of various scales, from a network of several dozen students to online networks with millions of vertices. In every scenario, we succeeded in identifying anomalous vertices with lower false positive rates and higher AUCs compared to other prevalent methods. Moreover, we demonstrated that the presented algorithm is generic, and efficient both in revealing fake users and in disclosing the influential people in social networks.