Collusive shill bidding detection in online auctions using Markov Random Field

Shill bidding is where spurious bids are introduced into an auction to drive up the final price for the seller. This causes legitimate bidders to pay more for the item in order to win the auction. Shill bidding detection becomes more difficult when a seller involves two or more bidders and forms a collusive group to commit price-inflating behaviour. Colluding shill bidders can distribute the work evenly among each other to collectively reduce their chances of being detected. This paper presents a Collusive Shill Bidding Detection algorithm to identify the presence of colluding shill bidders. The algorithm calculates an anomaly score for each bidder and then verifies the anomaly scores to improve the detection accuracy. We use a Local Outlier Factor for calculating the anomaly score for each bidder. We then model the auction network in a Markov Random Field and apply Loopy Belief Propagation for identifying the colluding shill bidders. We implemented the proposed algorithm and applied it on both simulated and commercial auction datasets. Experimental results show that the algorithm is able to potentially detect colluding shill bidders. Comparative analysis on simulated auction datasets shows that the proposed algorithm performs better than two existing published approaches.