With Fraud on the Rise, AI Can Fill in the Gaps
In today’s dynamic world of fraud detection, technology, and artificial intelligence (AI) are allies. The insights of industry experts, Yinglian Xie, a technology veteran with a background at Microsoft Research and CEO at DataVisor, Sandip Nayak, President at Fundation, and Andrew Davies, Global Head of Regulatory Affairs at ComplyAdvantage, discuss the transformative role of AI in fraud prevention.
When DataVisor started, it primarily offered advanced machine learning solutions, through an unsupervised approach. In other words, their programs can spot fraud without needing a loss or training labels; they can automatically identify suspicious activities. Xie explains that AI’s ability to make rapid decisions during real-time transactions depends on the amount of data available for this process. To achieve a proactive response, it must be synchronized with real-time data, as opposed to a manual or “supervised machine learning” approach.
“We need to kind of switch the traditional approach looking at fraud being very much kind of an isolated case, like a manual approach, and into something we need technology for, said Xie. “And we need to essentially be able to make decisions instantaneously as well.”
In addition to unsupervised learning algorithms, Xie explains that generative AI falls into another category of fraud detection. This method describes the data and communicates information back in human-like responses. Xie gives an example that as customers, some may not understand why a transaction was rejected and that’s where generative AI comes to rationalize the reason behind the rejection.
Echoing Xie, Nayak described solutions where traditional techniques fail, one of them being unsupervised learning algorithms. These algorithms can use techniques like anomaly detection to actually hone in on “the needle in a haystack problem.”
“Number two, the automated and advanced nature of AI can really solve the shortcomings of rules based and human based approaches in detecting fraud and can also self-calibrate itself as the nature of fraud evolves with time,” said Nayak.
Meanwhile, Andrew Davies pointed out that one of the biggest challenges faced by banks and financial institutions is “they are constantly playing catch-up.” With the accelerated pace of money movement and real-time settlement, he emphasized that fraudsters capitalize on this by being swift and innovative, continuously seeking out new vulnerabilities to exploit.
“Banks must update their legacy technology which leaves too many weak points in the control environment,” said Davies. “Additionally, as money moves more quickly and is subject to finality, fraud detection must be done in real time.”
And as the digital landscape continues to evolve, Nayak envisions the adoption of these technologies will be beneficial to the lending industry. Embracing different strategies not only reduces fraud losses but also enhances capital efficiency, paving the way for increased profitability and security in lending, according to Nayak.
“I do expect the lending industry, especially the ones who adopt the latest technologies of fraud detection, will have a competitive advantage compared to those who don’t,” said Nayak. “And what that will do is it will help them preserve more of their capital in the current tough macro environment by helping the overall unit economics…”
Unsupervised machine learning and generative AI are strategies reshaping fraud prevention. The ability to make rapid, data-driven decisions, adapt to evolving fraud tactics, and provide human explanations behind alerts has become a cornerstone in modern fraud detection.Last modified: October 19, 2023
Anaya Vance is a reporter for deBanked. Connect with me on LinkedIn.