A Multi-State Bayesian Network for Shill Verification in Online Auctions

Online auction systems have made remarkable progress in recent years. However, one of the most severe and persistent problems in such systems is shilling behavior, which is a type of auction fraud where a bidder artificially drives up the bidding price so that the winner of the auction has to pay more than he otherwise would pay. Verification of shill bidders in an online auction is difficult due to incomplete knowledge about suspicious bidders. In this paper, we introduce a novel approach for verifying shill bidders using a multi-state Bayesian network, which supports reasoning under uncertainty. We describe how to construct the multi-state Bayesian network and present formulas for calculating the probabilities of a bidder being a shill and being a normal bidder. To illustrate the effectiveness of our approach, we provide a case study for shill verification, and demonstrate that a multi-state Bayesian network performs better than a bi-state Bayesian network.