Mobile Payments Fraud Fight Gets a New Weapon in ACI Worldwide

March 10, 2020         By: Steven Anderson

Security has long been one of the biggest problems in mobile payments, and it’s kept a lot of otherwise potential users out of the fold. Improving security, therefore, has become one of the leading challenges to improving mobile payments, a challenge that has been met on several fronts and continues to improve. One new improvement on this front was recently rolled out by ACI Worldwide, who tipped us off about the effort.

The new improvement brought out from ACI Worldwide is a new kind of machine learning system, which can help detect patterns in customer purchases. The new “Incremental Learning” technology, as it’s known, is designed to not only spot patterns, but also improve on its ability to spot new patterns as they emerge. This helps ensure that the system won’t be overwhelmed by fraudsters changing their methods as they go.

Additionally, the system is also designed to flag these issues in real time, allowing for a more rapid response and with it better outcomes. The system is being incorporated into several ACI products, including both Proactive Risk Manager and ACI ReD Shield.

ACI Worldwide’s director of data science Jimmy Hennessy noted “Traditional machine learning models in many cases are not sufficient to stop fraudsters in their tracks. As fraudsters become more sophisticated, we need to continuously advance our models to beat them at their own game. Our global data science team has created a game-changing piece of machine learning technology that can be seamlessly integrated and future-proofs the precision and operational efficiency for over 5,000 institutions protected by our solutions today.”

The problem with tracking fraud cases like this is the potentially high number of false positives. If, for example, someone routinely shops on eBay or Amazon, and then one day places an order with a small beverage maker for some small-batch sodas, is that fraud, or is that just a shopper going somewhere different? Machine learning might flag the transaction as potential fraud, where there is no fraud. Still, false positives can be a useful thing; by questioning some transactions, fraud will likely be caught.

Machine learning isn’t foolproof, but it does learn. It’s right in the name. As long as it learns, it’s likely to improve and produce better results for the security platforms that report it. False positives will likely happen, and inconvenience many, but for those it protects, the ends might well justify the means here.