An early fraud detection mechanism for online auctions based on phased modeling

Abstract

Reputation systems provided by online auction sites are the only countermeasure available for buyers to evaluate a seller's credit. Unfortunately, feedback score mechanisms are too easily manipulated creating falsely overrated reputations. Therefore, developing an effective fraud detection method can assist the user in identifying cases of fraud. However, none of existing research addresses the most important issue of early fraud detection, which is, discovering a fraudster before he defrauds. For effective early fraud detection for online auctions, this paper proposes a novel phased detection framework to identify a potential fraudster as early as possible. To heighten precision in detection, different quantifiable behavioral features were extracted and integrated with regression model trees to build phased fraud behavior models. To demonstrate the effectiveness of the proposed method, real transaction data were collected from Taiwan's Yahoo!Kimo for training and testing. The experimental results with these models show that the recall rate of fraud detection is over 82%.