The Gately Report: Deduce Attacking AI-Driven Identity Fraud

Identity fraud is perpetrated every 22 seconds in the United States, and AI-driven identity fraud is giving bad actors even more of an edge as they rack up more victims.

That’s according to Ari Jacoby, CEO of Deduce. It detects “super” synthetic identity fraud, AI-generated identities realistic enough to fool legacy fraud solutions.

Deduce exposes fake customers using its technology and an identity graph that continuously observes, authenticates and analyzes the interactions of 840 million U.S. identity profiles engaging in more than 1.5 billion daily events across more than 150,000 sites and apps. It protects new account-opening workflows and exposes fake accounts already in a customer database.

Deduce's Ari Jacoby

“The only way to fight AI-driven fraud is with an AI-driven solution,” Jacoby said. “If you're going to have one lightsaber on this side, you better have one lightsaber on the other side. It's not going to work otherwise.”

How Synthetic Identity Fraud Works

Synthetic identity fraud, for example, is meant to trick financial services organizations, Jacoby said.

“What sophisticated bad actors will do is maybe take a baby's Social Security number or a dead person's Social Security number, or someone without housing, and they will mix it with other ingredients and bake a fraud cake with it,” he said. “They'll pick a real street address and a real name, and they'll kind of Frankenstein it together. And through trial and error, they will create a synthetic identity, a fake identity that purports to be real and might bypass some credit bureau checks or other checks that trap for specific personally identifiable information (PII). Once that identity is created and it gets into credit bureaus, the fraudster is good to go. The fraudster can be a repeat offender and can steal credit from multiple parties.”

Organized crime is increasingly committing synthetic identity fraud, Jacoby said. The threat actors are smart, talented, Ph.D.-level, computer science-driven operatives from places like North Korea, Iran, Russia and other hot spots across the world.

“It's unfortunately the most profitable crime in the world, and because it's such a profitable crime, these folks have the means to get access to Nvidia chips on the black market, for example, stolen or repurposed GPUs,” he said. “They’re building AI-driven programs to create synthetic identities using state-of-the-art, AI-driven techniques. And when they're successful, they create these synthetic identities that masquerade around the internet looking just like me and you.”

Busting Out

These threat actors will fraudulently obtain lines of credit and pretend to be a responsible debtor, paying it off and gaining higher credit limits, and eventually they’ll do what’s called “busting out,” Jacoby said.

“They steal everything,” he said. “And this is probably in the first inning of baseball. It’s already a bigger than $100 billion problem, $100 billion worth of economic violence. People will steal credit, but they also will break into accounts and just steal the money and move it to a place where there's no extradition, goodbye. It happens in the regular finance world. It happens in the crypto finance world. It happens anywhere.”

And anyone can be a victim of synthetic identity fraud, Jacoby said.

“Your likeness, your image, your voice print and your mannerisms are all over the place,” he said. “If you have a video on YouTube or if you've ever had an answering machine, or if your mobile phone features voicemail, your voice print is out there. If you've ever had even a static picture of you on the internet, like social media or otherwise, it's out there. So all of this stuff can be put into a variety of AI-based systems. Even in Adobe, I can create a walking, talking you with your inflection, with your mannerisms, the way you nod your head, the way you tilt your ball cap, the glare that's coming off of your glasses, the sheen on your nose, all these incredible details. You typically wear this kind of shirt and all of a sudden it's almost indistinguishable. It's a massive, massive problem because fraud just gets worse if two parties that are transacting believe that each is real and one turns out not to be.”

Scroll through our slideshow above for more from Deduce and additional cybersecurity news.