Payment fraud evolves in ways that are truly frightening — and with haste. Fraudsters are continually scrambling to keep up with, and have one foot in front of, technology, with recent stumbling blocks in the form of EMV, in the United States, likely to give rise to greater card-not-present fraud.
But the would-be payments criminals have potent weaponry at hand, including ever faster, ever more powerful and ever cheaper computing power, and they have been targeting, according to data science firm Feedzai, the weaker links that exist in the financial services chain.
In a recent whitepaper titled “A Primer to Machine Learning for Fraud Management,” the firm noted that, even as financial services evolve to embrace a digital world — with, say, virtual goods in hand and even virtual cash — the prospects for successful payments malfeasance grow in lockstep.
In fact, said Feedzai, as many as 65 percent of firms with annual revenues of at least $1 billion were victims of payments fraud as recently as 2014. That’s not all that far from the 56 percent of firms under that $1 billion top line threshold seen in the same year.
Against pervasive fraud, a vulnerable financial services industry and well-armed actors, what’s a firm to do? Companies cannot sit idly by, especially as the window to effective fraud prevention, detection and remedy increasingly becomes a smaller one. Acccording to Feedzai, there is one option: Machine learning.
Machine learning uses patterns within data that may not initially be readily apparent, with an emphasis on using artificially intelligent machines to dispense with rule sets and mine Big Data swiftly and completely — in a continuously refined manner and often in less than a second.
Certainly, the market is there for using machine learning to plow through electronic payments, looking for aberrant patterns. Feedzai noted in its research that of the $11 trillion in U.S. personal consumption expenditures, roughly 79 percent will be through electronic payments.
In an age where fraud prevention is still significantly a manual process, false positives abound, indicating wasted time and effort, even as the advent of new customer channels via online and remote payments activity increase the speed and severity of fraud itself.
Machine learning suits that urgent environment well, because it performs analytics and assigns risk scores in real time. There’s an added benefit in machine learning, called behavior analytics, which can help generate actionable ideas on how users can and likely will behave in given scenarios. The name of the game, according to the firm, is pattern recognition — but recognition across quite complex and subtle patterns. The use of machine learning can also be of benefit in driving new business, which, for a financial services outfit, could be useful across garnering new customers, performing underwriting and even scoring transactions at the point of checkout, possibly eliminating the need for chargebacks, helping financial results and engendering customer satisfaction.
. . . . . . . . . . . . . . . . . . . .
To learn more download the whitepaper “A Primer to Machine Learning for Fraud Management” by completing the below form.