TransUnion Announces Expanded Machine Learning Capabilities Within Device Risk Solution

TransUnion has expanded the machine learning capabilities in its Device Risk solution to improve fraud detection accuracy while reducing friction for legitimate customers.

What TransUnion announced

• TransUnion added new machine-learning models to its Device Risk solution, aimed at detecting more sophisticated digital fraud attacks across channels.

• The update is positioned as a response to rising digital fraud losses, which TransUnion quantifies at hundreds of billions of dollars in recent survey work with business leaders.​

Key new ML capabilities

• Stronger recognition of returning devices across different customers, helping link sessions and events even when identifiers change.

• More robust detection of non-human activity, including behavioral patterns associated with virtual machines, residential proxies, remote desktops and bots.

• Deeper consortium‑driven insights: the models learn from thousands of device signals plus fraud feedback from TransUnion’s global fraud consortium, improving responsiveness to emerging attack patterns.

How the ML works in practice

• Pre-built adaptive ML models are delivered as part of Device Risk, so clients do not need to build or train their own models to benefit.

• The models continuously retrain on confirmed fraud and device-level telemetry, allowing risk scores and rules to evolve dynamically rather than remaining static.

• TransUnion reports that these ML enhancements can improve fraud capture by up to about 50% compared with traditional, static, rule-based or device-recognition-only approaches, while also reducing rule-maintenance overhead.

Impact on fraud strategy and UX

• Higher-precision device risk scores allow organizations to step up authentication selectively (for risky devices) and streamline low-risk interactions, which can reduce customer friction at login, transaction, and account opening.

• By pairing enhanced device intelligence (fingerprinting, anomaly and evasion detection) with adaptive ML, the solution is intended to support real-time decisioning via APIs and integrate into existing fraud and identity workflows.