Highlights
AI is escalating the fraud arms race, enabling sophisticated tactics like deepfakes and synthetic identity theft. Traditional rule-based detection systems are insufficient, prompting a shift to AI-driven behavioral analytics.
Adaptive behavioral analytics (ABA) and automated deep behavioral networks (ADBNs) from Featurespace analyze hundreds of behavioral signals in milliseconds, allowing institutions like Visa to detect anomalies in real time and minimize both fraud and false positives.
Fraud prevention must evolve into a collaborative, ecosystem-wide effort, with financial institutions treating it as a customer trust initiative rather than a siloed security function.
Watch more: Visa and Featurespace Tackle Deepfakes as Fraud Fight Turns Predictive
In today’s war against financial fraud, the battlefield is evolving faster than ever, as artificial intelligence (AI) fuels both sides and the adversaries become increasingly sophisticated.
That’s why PYMNTS called on Visa DPS GM, Risk Products and Solutions Dustin White and Featurespace Chief Operating Officer Tim Vanderham to unpack how payments leaders are answering back with a pioneering approach grounded in behavioral analytics and collaborative innovation.
“It’s not about just stopping fraud that happened yesterday,” White said. “The real question is, are you prepared to stop the fraud that’s coming tomorrow?”
Fraud detection has long relied on rule-based systems that flag “bad” activity after it happens. While these systems have historically been useful, it’s becoming clear that they are no match for the generative AI-fueled tactics now used by cybercriminals like social engineering, synthetic identity theft and deepfake scams.
The new goal is to stop fraud before the transaction is even attempted.
“We’re not just looking at transactions,” Featurespace’s Vanderham said. “We’re looking at logins, payee data and behavioral metadata. The fraud event is often the final act in a longer play.”
To counter the generative AI-fueled tactics employed by cybercriminals, Visa DPS has partnered with Featurespace, a U.K.-based firm specializing in adaptive behavioral analytics, to refine its approach to fraud detection.
“Historically, fraud systems have focused on the bad,” said White. “But in a world of digital commerce, the importance of profiling the good truly can’t be understated.”
At the core of Featurespace’s model are adaptive behavioral analytics (ABA), paired with patented automated deep behavioral networks (ADBNs). These technologies enable AI systems to track a user’s normal behaviors — including how they spend, where they log in from, and which merchants they interact with — and flag deviations in milliseconds.
“Visa brings the breadth of global transaction data,” Featurespace’s Vanderham said. “We bring the depth of modeling and behavior profiling. One plus one, in this case, equals more than two.”
According to Vanderham, Featurespace models can analyze up to 500 behavioral variables in less than 50 milliseconds, cross-referencing new transactions with a user’s long-term behavioral history.
“If I’m making a debit payment at Starbucks every morning while I’m traveling, the system remembers that,” he said. “We take the transaction, parse it down to its components, and compare it against historical behavior at both the account and network levels.”
This immediacy matters more than ever. As digital payments become ubiquitous, even slight interruptions caused by false positives can send customers packing.
“Fraud isn’t just a financial exposure metric anymore,” White said. “It’s an experience metric. Miss one transaction or wrongly decline a legitimate one, and the customer might walk.”
The evolving nature of fraud is unsettling. Large language models can now impersonate voices, write believable scripts in any language and generate AI avatars indistinguishable from the real thing.
“Our own sales team received a deepfake of me asking for confidential information,” Vanderham said. “It’s no longer phishing. This is precision social engineering using voice, video, and data manipulation.”
Scams are becoming more humanized and emotionally manipulative. Upstream signals like login behavior, account origination patterns, and even typing speed, have an increased role to play in strengthening responses across earlier detection stages.
“Someone builds trust with a victim over days or weeks,” White said. “By the time they initiate the transaction, it looks valid. Stopping that at the transaction level is too late.”
Even more worrying is the rise of dormant mule accounts and cash-out schemes. One client on the U.S. West Coast used Featurespace’s tools to dismantle a fraud ring, resulting in 32 arrests and hundreds of thousands of dollars recovered.
“That’s what keeps me up at night,” Vanderham said. “Triangulated fraud, deepfakes and cross-border cash-outs that drain funds in minutes.”
“Fraudsters collaborate across borders, across verticals,” said White. “We have to be just as unified in response. It’s not a Visa problem, it’s not a Featurespace problem — it’s an ecosystem problem.”
Beyond technology, a cultural shift is also underway. Institutions must stop operating as isolated nodes and start thinking like interconnected networks. If there’s one immediate step every bank or credit union should take, it’s changing the conversation internally.
“Reframe fraud prevention as a customer trust initiative,” said White. “Get out of the fraud department mindset and make it an enterprise-wide priority.”
We’re always on the lookout for opportunities to partner with innovators and disruptors.
Learn More
