Detecting fraud in internet auction systems

Abstract

Fraud compromises the thriving Internet auction market. Studies have shown that fraudsters often manipulate their reputations through sophisticated collusions with accomplices, enabling them to defeat the reputation-based feedback systems that are used to combat fraud. This paper presents an algorithm that can identify colluding fraudsters in real time. Experiments with eBay transaction data show that the algorithm has low false negative and false positive rates. Furthermore, the algorithm can identify fraudsters who are innocent at the time of the data collection, but engage in fraudulent transactions soon after they accumulate good feedback ratings.