Fighting Fraud By Not Focusing On It

Machine learning can help find real cyber transaction fraud while letting the “good” payments go through. TSYS’ EVP Karim Ahmad and Featurespace CEO Martina King discussed their partnership geared toward battling the bad guys to reduce online fraud.

Fighting fraud is a lot harder online, and a lot harder for merchants and consumers, as card-not-present transactions become the preferred method of malfeasance.

In one recent announcement, payments provider TSYS and real-time learning technology platform Featurespace said they were joining forces to offer fraud prevention tools. The first product, the TSYS Foresight Score with Featurespace, utilizes machine learning and other initiatives to pinpoint cyberattacks tied to transaction fraud.

The stage is certainly set for new ways of thinking about and battling consumer fraud, as total losses from fraudulent transactions are set to cost consumers as much as $31 billion by 2020.

In an interview with PYMNTS’ Karen Webster, Karim Ahmad, executive vice president of Global Product and Innovation at TSYS, and Martina King, CEO of Featurespace, delved into the ways the companies are working together to limit both transaction fraud and the “false positives” that can stop a transaction in its tracks, when in fact money should be changing hands.

Said Ahmad, “The partnership is about using machine learning to identify and improve risk and fraud performance. This is an area where there is a lot of focus for issuers globally. We are always on the lookout for new technology that can get us better results” — a search that he said had led TSYS to Featurespace. And, as Ahmad told Webster, “a dollar that goes to a fraudster is a dollar that an issuer is not spending in other areas or on their programs.”

Featurespace traces its own genesis to academia, specifically Cambridge University, where the firm was spun out back in 2008. Under the guidance of Prof. Bill Fitzgerald, head of applied statistics and signal processing at that university, and his Ph.D. student David Excell, the nascent firm began its work separating signals from noise to determine what “normal” consumer behavior might look like to let legitimate transactions continue unimpeded by cybersecurity snares.

The firm’s first analytics system was, in fact, built to serve a peer-to-peer (P2P) gaming site. The overarching theme since then, as Martina King noted, has been to analyze “good” behavior rather than simply spotting anomalous behavior when examining banking and payments behavior.

Ahmad agreed, saying that focusing on good behavior “helps us drive down false-positives.”

In payments and in the application space, “there is a lot more noise … Having focus on what a good profile looks like for a consumer and then as that profile changes having access to the machine learning, to allow that profile to evolve … that is what is so powerful about this technology.”

TSYS Foresight Score℠, Karim Ahmad said, might be seen as “just dipping our toes in the water.” Moving forward, there will be time to “learn an enormous amount about how the technology scales, what it is good at and where it needs to be improved over the next few quarters as we fully get going with this product. The underlying technology is incredibly powerful.”

The TSYS executive told PYMNTS there would be an opportunity to tune scores and deliver better results, with the eventuality of scoring credit card and other applications (beyond just transactions) themselves during customer acquisitions.

Webster noted that there is no shortage of data depicting the mounting costs for firms as they implement fraud prevention tools. But faced with the goal of reducing cyberattacks or letting more customers transact, Ahmad stated that both endeavors should be in place.

Fraud attempts have moved toward card-not-present transactions, and so “we are never really done fighting transaction fraud,” said Ahmad. “The other part of it is that every good transaction that we let through is … a bad customer experience that was avoided,” especially when limiting the incidents of declines at the point of purchase, which can be fatal to the customer/merchant relationship.

It’s a predictable fact that consumers are going to be less predictable than ever when it comes to online transactions and as they do more and more business online. But King noted that granular information gleaned on each consumer, individually and instantaneously, can help gain a better sense of what buying patterns may emerge.

King and Ahmad agreed that their partnership will work together to, as Ahmad put it, “join up the information flow between the issuers and the merchants. That to me is a next area of innovation.”