A team of researchers has designed a way to identify vendors with bogus seller ratings.
To help combat bogus eBay (NASDAQ:EBAY) seller ratings, which can mislead customers by giving a stamp of approval to fraudulent vendors, researchers at Carnegie Mellon University have developed a technique to help validate the scores.
EBay fraudsters have found ways to artificially elevate their seller ratings, by conducting transactions with friends or using alternate online identities to post positive feedback about themselves -- much to the frustration of legitimate vendors who work hard to earn genuinely positive ratings.
"We want to help people detect potential fraud before the fraud occurs," research associate Duen Horng "Polo" Chau, one of the scientists behind the fraud detection technique, said in a statement.
Chau, along with professor Christos Faloutsos and two other students, analyzed roughly one million transactions between almost 66,000 eBay users. They plotted the transactions as a graph to identify the distinctive pattern created by sales between fraudsters and their accomplices, then designed an algorithm to detect unnaturally close-knit groups of people that traded mainly among themselves.
The method, dubbed "Network Detection via Propagation of Beliefs," or NetProbe, correctly identified 10 known fraud perpetrators, as well more than a dozen probable scam-artists and accomplices.
Although they have not yet released it to the public, the researchers developed a software program that provides a trustworthiness score for individual user IDs based on the NetProbe algorithm. The researchers hope their tool will enable shoppers to distinguish between legitimate eBay vendors and scammers.
However, Faloutsos said he recognizes the limitations of any reputation system. "Crooks are extremely ingenious," he said in a statement, adding that he hopes anti-fraud strategies like NetProbe will make it harder to manipulate reputation systems and make fraud "increasingly unprofitable."
Chau and his colleagues plan to publish a paper on their research early next year and some version of NetProbe may ultimately be marketed to eBay or directly to consumers.