Out-Thinking Hackers With Adaptive AI

Whether the objective is teaching behavior to kids, being an effective team member, or building more impenetrable cyber defenses, learning from a good example works.

This is pertinent to artificial intelligence (AI) in finance, where enhanced forms like adaptive AI are fulfilling the technology’s promise by actually “learning” — foremost from the behavior of legitimate deposit account holders — so that hackers stand out when they make a move.

Emerging uses of adaptive AI and other new tech is set out in the new FI Fraud Decisioning Playbook, a PYMNTS and Simility collaboration, as financial institutions (FIs) and their clients ready for reopening by modeling good customer behavior to help identify bad actors.

Learn by Example

The inaugural FI Fraud Decisioning Report takes a keen interest in asymmetries between good customers and cyberthieves, as the two groups’ data footprints are quite different.

“Fraud decisioning strategies are more effective when the data gathered and analyzed includes high-quality evaluations of legitimate customers,” the report states.

“These customers build long digital transaction histories across online and mobile platforms as well as through purchases made in stores with card and app payments. Artificial intelligence (AI)- and machine learning (ML)-powered tools can analyze this data to better determine legitimate users’ behaviors, and these advanced technologies can often catch small behavioral details that may elude even the most-talented human analysts. A clear fraud decisioning framework with robust data gathering and analysis can power FIs, protecting them from losing billions to fraud and helping them gain customers’ trust.”

Know your customer (KYC) requirements are the law, and the work mandated under KYC regulations is also valuable to FIs in getting to know good customers in the service sense. The insights gathered by adaptive AI and advanced machine learning systems come to understand customers as a set of patterns. Deviations from those patterns is a predictor of intrusion.

“You need insights into the data to see what looks like a normal transaction,” Wells Fargo Head of Merchant Services Colleen Taylor told PYMNTS. “Then, if anomalies start to stick out, you can anticipate that maybe something is wrong there.”

“The innovation in the fraud space is really fast because the fraudsters are really innovative,” Taylor said. “We’re trying to stay ahead of where the fraudsters are. They spend a lot of time and money on trying to cheat the customers, and we have to spend the same — if not more — on keeping our customers safe.”

Automate to Validate 

The data pools FIs can draw from, including payments flows in and out of deposit accounts over time, charge patterns, location and many other data points, are the raw material that AI and adaptive AI analyze to enable more accurate real-time decisions.

Reducing friction and false positives as more shopping moves into the digital domain post-pandemic are critical now. As hackers grow more sophisticated by the day, adaptive AI and data-driven decisioning platforms get to “know” real transactions from fraudulent.

“Sophisticated automated tools have grown increasingly crucial to protecting banking operations as the number of fraud events utilizing stolen or falsified information grows,” the report states.

“FIs are exploring both AI and ML to protect daily transactions — often made in near real time — across multiple markets and networks in quantities that would likely be impossible for human employees to validate without the help of automation. Automated tools can recognize data patterns much more quickly than humans and alert employees to potential fraud.”