MasterCard International Inc. has commissioned a government-sponsored research laboratory to develop advanced antifraud technology.
The card association said last week that it expects the arrangement, with Los Alamos National Laboratory in New Mexico, to bear fruit within a year in the form of pilot tests of systems that can be of exclusive benefit to MasterCard members.
The aim is to go beyond the neural networks that have become increasingly popular in credit card fraud prevention, said Philip P. Verdi, MasterCard's executive vice president of electronic services.
He said MasterCard is less concerned with one-upping Visa, which has neural network programs well under way, than it is with providing systems tailored for MasterCard's more decentralized operation.
Drawing from the scientific disciplines available at Los Alamos, MasterCard expects to "complement the technology that many of our members have invested in or benefit from through third-party processors with neural networks in place," Mr. Verdi said.
Steve Coggeshall, a Los Alamos physicist serving as MasterCard project manager, said most of the banking industry is "narrowly focused on standard neural network technology." His group will "work with a variety of technologies that, in combination, will provide a cutting-edge, unique solution for fraud prevention."
Among the new trails he expects to explore are fuzzy logic, which is designed to approach the abstract thinking that is beyond conventional computer programming; and genetic algorithms, which simulate evolutionary processes to arrive at optimal solutions to problems.
Neural networks in banking have become almost old-hat. A product of artificial intelligence research, neural systems are programmed to "learn" based on human-brain-like structural principles.
They have proved especially useful in recognizing anomalous patterns within massive quantities of data that otherwise would require great time and effort to pore through and analyze.
Credit card banks have reported significant fraud-loss reductions after subjecting transaction data to neural pattern analysis. HNC Software Inc. has led the commercialization of neural technology for the card industry, working with most of the major card-issuing banks, processors, and with Visa U.S.A.
San Diego-based HNC also has gone beyond basic consumer account and transaction analysis into merchant risk analysis to help banks and merchant processors ferret out money laundering, skimming, and other improprieties at the store level - an area that MasterCard and Los Alamos also intend to concentrate on.
Mr. Verdi said eight of MasterCard's top 10 global card issuers have their own neural networks. "We have therefore decided to focus our efforts on modeling data that issuers do not regularly see, such as merchant locations and transaction category codes," he said.
Whereas banks have sufficient cardholder data to develop their own customer risk scores, MasterCard and Los Alamos said they will "create a predictive model based on a more comprehensive set of data and criteria."
For Los Alamos, which traces its history to the atomic bomb program in World War II, the MasterCard contract is a technology transfer to the private sector - a post-Cold War priority for all of the Department of Energy's research labs.
For MasterCard, Los Alamos will draw on technology that has been used to detect nuclear weapons proliferation and search archival data bases from weapons tests.
Los Alamos does other corporate work, including an extensive, jointly funded research and development program with Citicorp. With other labs it has participated in the Financial Services Technology Consortium, a group of banks organized by Citicorp to explore leading-edge technologies.
"The lab applies problem-solving in a unique way, at times not even knowing if there is a solution," Mr. Coggeshall said. "We'll try many techniques to find out what works and doesn't work, which contributes to invention . . . We're doing in the industrial world what we have been doing in the scientific community for the last 10 years."
