8.8 Auction Fraud

The most damaging act that can be perpetrated against an on-line auction site like eBay.com is to tarnish its visitors' trust of the site. On-line auctions like eBay, Yahoo, and Amazon have started to attract situations where successful bidders pay for items, such as a camera or a laptop, and then never receive the goods after mailing their money. Interestingly, the large auction sites disclaim most of the responsibility for combating fraud that takes place in their marketplaces. For example, eBay says fraud accounts for a minuscule portion of its auctions—citing that less than 0.1%, or 1 out of every 40,000 listings, result in a confirmed case of fraud. Furthermore, it claims most items are covered by insurance for up to $200, less the $25 deductible. Despite its insistance that fraud is not a major problem, eBay created a new data mining unit designed to detect instances of crimes. It also developed a system called the Fraud and Abuse Detection Engine (FADE) to detect perpetrators at its site.

The Consumers League estimates most fraud takes place with large-ticket items that are not covered by insurance. Statistics from fraud.org confirm the magnitude of the on-line auction problem (see Table 8.1).

Table 8.1: On-Line Auction Fraud Statistics
2000 Top 10 Frauds		Jan.-Oct. 2001 Top 10 Frauds
On-line auctions	78%	On-line auctions	63%
General merchandise sales	10%	General merchandise sales	11%
Internet access services	3%	Nigerian money offers	9%
Work-at-home	3%	Internet access services	3%
Advance fee loans	2%	Information adult services	3%

Ruben Garcia, assistant director of the FBI, defines Internet fraud as "Any instance in which any one or more components of the Internet, such as the Web sites, the chat rooms, e-mail ... play a significant role in offering nonexistent goods or services to customers, in communicating false or fraudulent representations about schemes to consumers, in transmitting victims' funds or any other items of value to the control of the schemes' perpetrator."

Obviously, this can also be interpreted as including claims made about goods up for bidding in an on-line auction, which after payment is received, are never sent to the successful and unsuspecting bidder. Most large auction sites take some level of security to ensure that the identity and reputation of those participants in their site are legitimate and trustworthy; however, given the volume of bids placed on a daily basis, perpetrators are going to make their way into the marketplace.

Most victims of Internet auction fraud who contacted the FBI were males between ages 20 and 50. The most common frauds involved Beanie Babies (27%); however, more expensive items were also high on the list of scams, including video consoles, games and tapes (24%), and laptop computers (18%). The average victim lost $776, with most victims paying by money order or check and who typically only knew of an e-mail address or post office box number, according to the FBI. Of an estimated 500 million Internet auction sales last year, the bureau estimated about 5 million were fraudulent.

The biggest form of on-line auction fraud is where a seller takes payment and then fails to deliver the promised item. Alternatively, the item is delivered, but doesn't match the items description at the auction site. A seller might use flattering or deceptive photographs and "optimistic" descriptions, for example, to pass off damaged or second-rate goods as high in quality. In many cases, a mismatch between the item described in the auction and the item delivered may simply be a question of interpretation; one person's "good condition" may be another's "average condition." Sellers may overcharge buyers for packaging and shipping fees. Unscrupulous sellers are happy to make their money on inflated shipping charges for which they don't need to pay a percentage fee to the auction site.

One of the ways many auction sites attempt to regulate sales is through a feedback mechanism. When an auction is over, the winner/buyer can grade and comment on the seller and the transaction, and all grades and buyer comments are published for all to see. Obviously, buyers are more comfortable bidding on items from sellers who have received a lot of positive feedback and high grades.

Inevitably, then, sellers may attempt to fix their grades by submitting feedback about themselves or by getting friends and confederates to provide glowing reports. Fraud is not all one-way of course. Buyer collusion is where one buyer might make a low bid, and a second buyer then immediately makes a very high one, thus ensuring that nobody else makes any other bids. At the last minute, the second bidder retracts, allowing the first bidder to get the item for a very low bid.

The ability to disguise identity, revoke bids, and maintain multiple on-line identities may facilitate undesirable practices like shilling. Shilling is where sellers arrange for false bids to be placed on the items they are selling. Sellers place the bid themselves by using multiple identities or by using confederates. The idea is to force up the cost of a winning bid and encourage interest in the auction.

More stringent control over multiple identities and an analysis of the types of goods with a higher percentage of being fraudulent by these auctioneers could reduce the likelihood of criminal action on their site. Once an individual has perpetrated a fraud, an investigation into his or her identity, physical address, P.O. box, e-mail, IP address, and Internet provider should be aggressively pursued. Once identified, all this information should be collected in a centralized database in which queries from all major auction sites can be processed prior to the completion of auctions. Certain items, particularly dollar ranges, tend to attract fraud. These attributes, goods, and intervals are capable of being identified via data mining analysis. In the end, the solution to auction fraud is a combination of common sense, centralized access to a database of known criminals, and data mining models. As we have seen with eBay, concerns about protecting its bidders forced it to create a data mining Security and Trust Group to stem the flow of auction fraud.