Mercator Advisory Group recently released a new study entitled “Five Technologies That Will Impact Payments and Financial Institutions Going Forward,” and machine learning has topped the list.
While the predictive analytics delivered by machine learning have long been leveraged by neural networks to detect fraud, to date even the most advanced systems have required manual programming in order to perform.
However, a new generation of machine learning technology is quickly emerging for fraud detection – and it is faster and smarter than ever. And the ability of these tools to dive deep into transaction data, pinpoint anomalies, and make instantaneous decisions about fraud as it unfolds is projected to save our industry billions.
In fact, research from Oakhall and published on finextra.com estimates that financial services firms worldwide could save at least $12 billion a year by optimizing adaptive, machine learning fraud management technology.
And to bring the far-reaching benefits of machine learning to its client credit unions, CO-OP Financial Services is making significant investments in the technology this year.
CO-OP’s Machine Learning Platform
According to Nathan Rogers, senior manager, product marketing for CO-OP, a new machine learning solution is scheduled to go live in 2017 that will build on the company’s existing advanced fraud detection systems, anchored by the FICO Falcon Fraud Manager.
“The Falcon Fraud Manager is widely recognized as the industry standard today for detecting fraud, and the addition of our new machine learning platform will only strengthen our transaction scoring model,” he said. “In optimizing both technologies, we expect to reduce fraud losses significantly while keeping false positives to a minimum.”
Machine Learning and Neural Networks
Rogers notes that a major new benefit machine learning will bring to CO-OP is its ability to rapidly adapt decision making.
“Neural network platforms detect, flag and block transactions according to fraud strategy rules that are already in the system and that apply to known threats,” he said. “Machine learning will allow us to detect new threats immediately, block them at the exact moment that the fraud is occurring – and automatically apply the data to improve the model’s future performance.”
Machine learning will also bring to CO-OP the ability to block fraud on a more individualized basis.
“Sometimes we find that fraud is happening to a single credit union member, which means there is no reason to update the rules system-wide,” he said. “Machine learning technology accurately identifies fraud at a more granular level based on historical data in the system, without the directive of a fraud strategy rule.”
And, Rogers emphasizes, one of the most valuable benefits of machine learning is the ability to feed numerous external data sources into one engine in order to build more robust models, catching more fraud faster.
Beyond Card Transactions
While machine learning promises to dramatically curb card fraud, account-based transactions also require fraud protection.
“Our machine learning solution will protect account-based payments as well,” said Rogers. “Frequently fraudsters are compromising cards and accounts at the same time, a scenario that can be more difficult to catch because criminals vary their activity in order to avoid detection. Machine learning will provide us with a platform that analyzes both types of fraud in a unified manner, allowing us to battle fraudsters more effectively on all fronts.”
Rogers adds that detecting account-based fraud is becoming increasingly important in the age of real-time payments.
“It used to be that all account-based payments went through the ACH network, which meant there was time to consider whether a transaction was fraudulent,” he said. “With new payment types and real-time channels going mainstream, monitoring these payments and their potential risk in real time is more important than ever.”
He continued, “Remember also that MasterCard and Visa only offer chargeback rights for signature transactions. So PIN-based fraud detection is also a critical component to a successful fraud mitigation strategy. Credit unions need to closely watch all payment types in order to protect members well. Ultimately, our goal is to strengthen every payment that comes in and out of the credit union, and to safeguard member card and account data with an all-encompassing approach to security.”
The original article Machine Learning: Real-Time Fraud Detection Across Cards and Accounts can be found on Insight Vault.