The advent of online platform economy has increased online fraudsters. To detect them, most research explored to recognize characteristics from behavioral data of online users. In those works, each user was regarded as an individual subject, while relationship between users was underestimated. This paper introduced methodology of social network analysis to mine characteristics of relationship among online fraudsters. Using dataset from a Bitcoin trading website, a weighted signed network is constructed. By removing normal user nodes from the network, relationships between fraudsters are clarified. Then, major structural patterns of the snipped network are uncovered by way of community partitioning method. Based on real data from the Bitcoin trading website, 3 typical structures of fraudster groups are found: star-shaped structure, double-core structure and reticular structure.
Hu, Jinya; Li, Feng; Wang, Yanfeng; and Zhuang, Dong, "Structural Pattern Recognition of Fraudster Groups in P2P Transaction Websites" (2018). ICEB 2018 Proceedings. 38.