Detecting Collusive Fraud in Social Network of Online Auction

During last years, online auction has attracted many researchers. However, growing popularity in online auctions result in increasing fraud in online auctions. For instance, fraud increased in eBay, as a popular company in online auction. Therefore, many researches have done for detecting fraudulent buyers and sellers. One of the fraud type in online auctions is collusive auction fraud, in which multiple seller and bidders collude with each other. This kind of fraud is dangerous and caused catastrophic financial losses. Therefore, many techniques proposed to deal with this kind of fraud in online auctions. In this paper, we propose a novel detection technique in online auctions that use one-class to calculate an anomaly score for each unlabeled user. Then it models the users’ interactions in the auctions as a pairwise markov random field (MRF). Next, our technique applies belief propagation to the MRF to revise anomaly scores. The results of our experiments show that our proposed technique is able to detect different types of collusive auction frauds within a reasonable detection time.