Internet Auction Fraud Detection Using Social Network Analysis and Classification Tree Approaches

Effective Internet auction fraud detection has become an emergent issue in real-world scenarios in conjunction with the rapid development and prevalence of Internet auctions. Accurate detection of fraudsters can assist law enforcement agencies in preventing potential fraud cases that could result in large monetary losses. This paper proposes a hybrid approach utilizing network metrics and data-mining techniques to discover fraudsters based on Internet auction transaction records. Using experimental data gathered from the Yahoo! Auctions Web site, extensive experiments demonstrate that the proposed approach is capable of detecting Internet auction fraudsters both effectively and almost instantaneously with an acceptable classification accuracy rate. The research findings will assist in probing the implications regarding different transaction account types, and they also reveal the specific transactions within various transaction patterns.