TransUnion LLC has introduced a major upgrade to its Device Risk fraud-detection platform, adding new capabilities designed to improve device recognition, identify anomalies and apply adaptive machine learning in real time.

The company said the enhancements provide businesses with clearer visibility into risky digital behavior as well as improved clarity on legitimate transactions.

Citing its own research, which found that businesses lost an average of 7.7% of revenue to fraud in the last year, TransUnion said the new capabilities should be part of a multilayer approach to fraud prevention that includes identity verification, behavioral analytics, risk scoring, rule-based decisioning and transaction monitoring.

Despite improvements introduced by artificial intelligence, fraud prevention remains a cat-and-mouse game, said Clint Lowry, TransUnion’s head of global fraud strategy.

“Fraudsters are always looking for vulnerabilities,” he said. “As we put in new detection capabilities to stop what they’re doing, they continue to search around for the next weak point.”

Device Risk is a passive technology that collects data “behind the scenes without the customer actually having to take any overt action,” Lowry said. The system analyzes browser or mobile characteristics to determine whether a device has been seen before and whether it exhibits suspicious activity.

De-emphasizing cookies

Lowry said the upgraded platform is intended to reduce dependence on cookies as a primary means of tracking a device. Third-party cookies are small pieces of data that a website stores on a connected device to recognize it on future visits. Some major browsing platforms are limiting or eliminating the use of third-party cookies for privacy reasons.

“We do still use cookies,” Lowry said. “We just don’t want to be dependent on them. Fraudsters routinely delete cookies to obscure their identities, so TransUnion has developed other proprietary mechanisms to identify returning devices.

Lowry declined to reveal the specific tactics Device Risk uses but said the software’s detection algorithms generate risk scores from thousands of device attributes and signals. Its probabilistic model evaluates what he called “unique or semi-unique factors” to determine whether a device matches one previously observed. “We collect thousands of data elements from the device and essentially run an algorithm that looks for various patterns,” Lowry said.

A major component of the update is an adaptive machine learning model that learns from new data automatically. TransUnion has used machine learning for fraud detection for years, but the new model has shown significant improvement.

IP intelligence

“We’ve seen clients that improved their fraud detection rate by 50%,” he said. The new software also integrates the IP Intelligence database TransUnion acquired with the purchase of Neustar Inc. in 2021. Lowry called it “the most robust database of intelligence around IP addresses on the market today.” IP Intelligence allows businesses to filter out trusted traffic more quickly, saving time at a critical point when customers are about to complete a transaction.

Lowry acknowledged that even the best technology offers no guaranteed protection against fraud. Artificial intelligence has given fraudsters powerful new tools to generate synthetic identities and deepfakes. “If we want to look at this objectively and say who’s winning this war, the fraudsters are winning,” he said.

That doesn’t mean business owners should despair. Lowry said TransUnion’s realistic objective isn’t to eliminate fraud entirely but to strike a balance between preventing it and allowing legitimate transactions to proceed without friction. “The more barriers we put in front of a consumer, the more likely they’re going to go to a competitor,” he said. “People need to transact.”

Photo: TransUnion