Presented by AU10TIX
Generative AI can write a sonnet in five seconds, but it can also cause billions of dollars in identity fraud and profoundly damage reputations of individuals, organizations and governments in an instant. AU10TIX’s Q2 2023 Global Identity Fraud Report warns that organized ID fraud has surged by 44% in the U.S., compared to preceding quarters.
AU10TIX’s report data comes from its AI-powered fraud detection tools, especially Serial Fraud Monitor. This tool has detected both an upswing in professional organized attacks and a broad array of attack modes across the globe, most notably in the U.S., and largely in the vulnerable crypto trading and payments verticals. Like many fresh challenges in a modern society, this upsurge is a direct consequence of living in an increasingly digital world. Trust between financial services, the consumers who use them and the technology that powers their secure interactions is growing more difficult to establish every day – and gen AI is profoundly complicating the issue.
“Large language models and generative AI give professional fraudsters the ability to mimic valid identities in convincing ways, so differentiating between forgeries and the real thing has become a major challenge,” says Dan Yerushalmi, CEO at AU10TIX.
“Professional fraudsters aren’t opening one account or stealing a few thousand dollars,” adds Ofer Friedman, chief business development officer at AU10TIX. “These are major swarm attacks that can last weeks, even months, and their impact is much bigger. Organizations are enduring harder blows, and the general response we see is helplessness.”
And consumers increasingly demand security and control, even over convenience. Organizations that can’t provide that level of security are losing customer faith.
The role of generative AI in fraud
“Generative AI puts all the data in the world at the fingertips of fraudsters, and can infinitely iterate to overcome thwarted attacks or simply cause more damage,” Yerushalmi says. “It is both simple and inexpensive. As a result, it’s attracting new players, who pay almost nothing for an array of cyberattack tools and produce significant results.”
Gen AI is impacting fraud in multiple ways. For instance, deepfaked images, voice recordings or videos can be used in advanced social engineering and phishing attacks. Deepfakes can also be used to create fraudulent accounts with fake identities or bypass ID verification to break into existing accounts. Fraudsters are even defeating biometric ID validation systems with presentation attacks, which use deepfakes to mimic a real human in a scan, and injection attacks, which take over the system’s camera to deliver an image directly.
And generative AI isn’t just manipulating existing information – it’s manufacturing entire identities.
“It turns out that fraudsters are better adopters of AI than anyone else,” says Friedman. “Generative AI enables fraudsters to change the rules of the game. Up until now, detecting fraud was about identifying signs of manipulation. But that approach is fast becoming useless, because now fakes are created from scratch.”
Synthetic IDs that blend stolen Social Security numbers, names and addresses with false information and new digital alter egos can now pass almost any verification check. Over time, they develop digital personas by mimicking real human behavior, which makes them even more difficult to detect. This sets the stage for a coordinated attack that could include credit fraud, account takeovers, money laundering schemes and more.
Turning the tables with AI tools
How do you fight fraud when it’s become essentially undetectable? The answer, Friedman says, is to add a layer of generative AI detection capabilities on top of your current security posture, and go deeper.
“To stem the tide, you need to add another line of defense,” he explains. “Instead of the usual single-layer case-based fraud prevention, we are moving into multilayer detection, which looks at fraud and the behavior of fraudsters from different angles.”
AI deepfake detection uses algorithms to determine if content has been manipulated, analyzing elements like facial transformations, skin texture discrepancies, unnatural eye movements, frame-to-frame lighting changes and more. Biometric markers such as facial landmarks, voice patterns and behavioral signals add another layer of scrutiny to differentiate between real and manipulated content.
Organizations should implement automated AI and ML identity verification systems at the case level, and add a second layer of defense that detects sophisticated coordinated attacks at the network traffic level. This requires an advanced neural network that can analyze subtle fraud patterns and behaviors in live traffic moment-by-moment to enable real-time intervention. AU10TIX’s traffic-level fraud detection solution, Serial Fraud Monitor, also pools anonymized data from dozens of institutions with large customer bases, which helps pin down hidden patterns and anomalies that point to widespread sophisticated fraud.
Yerushalmi adds that, although gen AI is going to increase the capacity of both fraudsters and fraud detection, it can’t do everything.
“We need to be careful not to put all our eggs in that basket and forget the basic algorithms handled by traditional software, because fraud comes from every direction,” he says. “An intelligent combination of machine learning, AI and basic algorithms is the key to optimizing results.”
And optimized results are crucial because, although today’s users are more concerned with security and control, they still demand a smooth and user-friendly experience.
“Customers will abandon any business process they are not happy with and move on to any of a thousand other options at their fingertips,” Yerushalmi says. “Users lose patience within five to 10 seconds. Adding these tools and technologies to the authentication process means greater protection that’s also seamless.”
An obligation of trust for customers
Protecting customers from fraud and scams is not only an organization’s most important legal and ethical responsibility, it is vital to the survival of the business, says Yerushalmi. It’s about protection from financial loss, of course, but also about building and maintaining customer trust and loyalty. The collaborative effort of organizations around the world is the only way to successfully combat widespread sophisticated fraud and retain consumer trust.
“Organizations must work together to reduce fraud globally, and AU10TIX brings a long history of relationship and trust building to that effort,” he says. “As new technologies emerge and gen AI continues to evolve and create new threats, we’ll continue to serve as a beacon of trust in today's increasingly digital landscape.”
AU10TIX’s Serial Fraud Monitor is an industry-leading solution that uses advanced neural network technology to provide businesses with real-time protection against sophisticated organized ID fraud, including swarm attacks and synthetic fraud. Learn more and schedule a demo here.
Sponsored articles are content produced by a company that is either paying for the post or has a business relationship with VentureBeat, and they’re always clearly marked. For more information, contact sales@venturebeat.com.