Research

Technological Possibilities – Machine Learning for Smarter Payment Security

By Beth Kotz, Aug 01, 2017

The ever-increasing role of technology in the modern marketplace has made transactions quicker and more convenient than ever for both businesses and consumers. Unfortunately, it’s also invited a greater risk of payment fraud and other cybercrime. Fraudsters have access to a variety of sophisticated attacks that can cause tremendous harm in a very short period of time – a reality that requires advanced tools capable of rapidly predicting, detecting and responding to suspicious activity and adapting to a constantly evolving digital landscape.

Perhaps no such tool is more powerful than machine learning, and businesses are increasingly turning to this technology to guard themselves against cyberattacks.

The Reality of Modern Payment Fraud

The concept of payment fraud is certainly nothing new. At its most basic, payment fraud is any attempt to deprive a victim of their funds, possessions or other sensitive data, often through means such as phishing, identity theft or merchant identity fraud. The financial toll of such a scheme can be devastating, but it’s far from the only risk.

Modern payment fraud can also destroy public trust in a company, create a litany of legal issues, incur many tangential costs and even open the door to account takeovers and theft of other sensitive data. Worse still, payment fraud is increasing, with nearly three-quarters of companies experiencing one or more fraud attempts in the last year alone.

Security through Data Analysis

To tackle this vital issue, businesses and other financial institutions are seeking new ways to leverage the extraordinary wealth of data available to them. Machine learning, in particular, is invaluable for sifting through vast stores of data, both historical and real-time.

One of the most important developments is the ability to scan years of purchase histories in seconds, which can be used to develop patterns and trends that establish “normal” behavior. Both supervised and unsupervised machine learning can be used to this end, creating sets of patterns that constitute routine behaviors for both large groups and individual consumers. Once normal behaviors have been established, computer algorithms can be used to assess individual transactions against typical patterns and identify potentially suspicious activities.

This approach is often paired with social network analysis, which expands the scope of fraud detection by seeking connections between individual actors across many channels and networks. Because fraud is often committed by groups of people rather than independent actors, social network analysis helps to uncover connections and detect a wide variety of potentially fraudulent activities more quickly and more broadly than other detection methods.

Gathering Data to Determine Probability of Fraud

As the velocity of commerce continues to intensify, machine learning models are able to distinguish subtle patterns and autonomously update to keep pace with what’s currently trending. The “self-learning” capabilities of these systems grow more effective with increasing data sets. Armed with a highly accurate set of training data for selecting the probability of a genuine transaction from a fraudulent one, a machine learning algorithm will be trained to pick up probable instances of fraud in the future.

However, human insights are still needed to produce a good machine learning model. PayPal for example has turned to a “homegrown artificial intelligence” built with open source tools to lead its cybersecurity approach, resulting in an impressively low fraud rate accounting for just 0.32 percent of its revenue. This compares very favorably to the average rate of 1.32 percent for all merchants.

Machine Learning in the Financial Sector

Though machine learning has found a powerful application in payment fraud prevention, its models are also used in investment, prompting the rise of so-called “robo-advisors” capable of effectively managing portfolios based on investors’ ages, financial statuses and other factors. Lending and insurance underwriting, too, have gone digital. Machine learning algorithms have proven effective at detecting relevant trends, assessing risk profiles and completing other tasks that previously required dedicated hours of human labor.  

Artificial intelligence is also increasingly being used in the credit card industry to produce more accurate transaction approvals and reduce the number of false declines. MasterCard, in particular, has led the charge with its Decision Intelligence platform, which uses machine learning to cut down on the approximately 15 percent of consumers who experience false purchase declines on legitimate purchases. The rise of mobile payments, as well as the proliferation of devices like Amazon’s Echo and Google’s Home that build databases of interactions with humans, has created an explosion of data that can be mined for a variety of useful insights.

As the financial and consumer marketplaces move progressively into the digital space, payment fraud and other criminal activities will only become faster and more sophisticated. In order to keep up with this growing and evolving threat, companies must use every tool at their disposal to fight fraud and keep their customers – and themselves – protected. Machine learning has already proven to be a valuable asset in this battle, and its importance will likely only grow in the future.

About the author

Author photo

Beth Kotz is a freelance writer and contributor for numerous home, technology, and personal finance blogs. She graduated with BA in Communications and Media from DePaul University in Chicago, IL, where she continues to live and work.


Tags

Unbiased Actionable Insights

Accelerate your organization’s journey to analytics maturity

Get the data sheet to learn how the Research & Advisory Network advances analytics capabilities and improves performance.

Download data sheet »

Become a RAN Client

Get answers to your toughest analytics questions with IIA's Research & Advisory Network.

»