Fingerprint, a U.S.-based device intelligence platform for preventing fraud, has introduced AI-driven recommendations to its Suspect Score solution.

This enhancement aims to empower fraud teams by enabling them to train machine learning models using their own labeled fraudulent data. This will generate optimized signal weights that align with their unique traffic patterns rather than relying on static and manually configured scoring models.

Built on Fingerprint’s suite of Smart Signals, real-time device intelligence insights, Suspect Score is utilized by enterprise fraud and security teams to evaluate transaction risk. The new AI-enhanced layer allows customers to upload their labeled fraudulent data for model training. Subsequently, it analyzes this data along with the Smart Signals to adjust signal weights, minimize false positives, and keep detection accuracy current as fraud trends change.

Customers can preview all recommended changes before applying them with a single click, maintaining full visibility and control over their scoring configurations.

Addressing the limitations of static models and adaptive detection

Static fraud scoring models often struggle to adapt to dynamic, traffic-specific fraud patterns that differ by business and evolve continuously. Fraud teams frequently lack the resources and time needed for manual analysis of signal interactions and retuning model weights specific to their use cases. Fingerprint asserts that AI-powered recommendations handle this process autonomously based on each customer’s data.

The update also addresses the growing complexity posed by sophisticated AI-driven bots and agents capable of circumventing traditional detection methods, as well as the rising use of privacy tools like VPNs among legitimate users, which makes conventional signal weighting less effective.

AI-powered Suspect Score recommendations are now available to all Fingerprint customers with access to Smart Signals. They can be activated through the Fingerprint dashboard.

Commenting on the launch, Valentin Vasilyev, CTO and co-founder at Fingerprint, emphasized that fraud patterns vary by business and constantly evolve. He noted that AI-driven recommendations eliminate the manual tuning bottleneck by training models with each customer’s labeled data.