|
Abstract: |
Shilling behavior is one of the most serious fraudulent problems in online auctions, and it is very hard to detect. Shilling is usually used to make winning bidders pay more than they were supposed to for the auctioned items. Since shilling behavior is not common in traditional auctions that are not traded online, this problem did not come to people???s attention until recently. As a result, there is a lack of effective mechanisms to detect, predict and prevent such fraudulent behavior in online auction houses, like eBay.
In this thesis, we propose a formal approach to detecting shilling behavior in concurrent online auctions using model checking techniques. We first developed a model template, which is based on the specification language of a model checking tool called SPIN. The model template represents two concurrent auctions. After applying real auction data into the template, it becomes an auction model, which simulates the bidding process of the two auctions. Then we use LTL (Linear Temporal Logic) formulae to specify normal and questionable behaviors. Finally, we use SPIN to formally verify if a bidder does shill in the model. To facilitate the entire process, we also developed a front-end tool to provide users a graphical interface.
This approach simplifies the problem of searching for shilling behavior in concurrent online auctions into a model checking problem. It can effectively detect potential shill bidders. To illustrate the feasibility of our approach, we have also performed some experiments using real auction data collected from the eBay.
|
