Summarize this article with
Fraud doesn’t stand still. Patterns shift, attacks evolve, and what worked last month may no longer apply today. Most fraud teams are building their defenses from rich datasets: device intelligence, behavioral indicators, and risk signals that help identify suspicious activities.
But the challenge isn’t just having the right data and the right signals. It’s knowing how to combine them effectively. Until now, teams have had to manually tune signal weights, adjust configurations over time, and rely on intuition and experience to decide what matters most.
To address this, we are excited to introduce AI-powered recommendations for Suspect Score. You can now use your own labeled data to generate a single, optimized signal weighting that improves fraud detection accuracy while maintaining full transparency and control.
Why fraud scoring is hard to get right
Every business has a different fraud profile. Attack patterns vary based on user behavior, geography, product design, and incentives. A signal that strongly indicates fraud in one environment may be much less useful in another.
Suspect Score already combines multiple signals into a single score, with default weights based on global detection rates. This is a strong starting point, but it may not always be the best fit for your specific traffic mix. Tuning those weights has always required manual work based on intuition or limited analysis. As fraud patterns evolve, these configurations can quickly become outdated or misaligned with what is actually happening in production.
The result is that it becomes harder to respond to threats with speed and confidence. And harder to optimize detection over time.
What’s changed: AI-powered Suspect Score recommendations
Suspect Score now introduces a new capability: training your scoring weights directly on your own fraud data.
Instead of manually guessing how signals should be weighted for your environment, you can now upload labeled fraud data and let machine learning generate an optimized configuration tailored to your traffic.
The system uses Smart Signals and your labeled outcomes to learn which signals matter most and how they should be combined, returning recommended weights that better reflect real-world fraud behavior.
Each time you upload new labeled data, Suspect Score weights can be recalculated to reflect your current fraud patterns.
From data to decision, with full transparency and control
The process is designed to be simple and transparent.
You upload labeled fraud data directly to the dashboard. Once your data is processed, we generate recommended weights for each signal based on your labeled fraud data, showing how each weight should increase or decrease. From there, you can review the recommendations and decide whether to apply them or keep your existing configuration. Nothing changes unless you choose to apply it.
At the same time, nothing becomes opaque. You maintain full visibility into how scores are constructed and full control over whether and how changes are applied. Suspect Score combines ML-driven optimization with transparency and ownership, allowing your detection to adapt as fraud patterns evolve without adding operational complexity.
Get started with Suspect Score
Your fraud data already contains the information you need. Upload it to the dashboard and get optimized weight recommendations tailored to your traffic with no manual analysis required.
AI-powered Suspect Score recommendations are available to all current customers using Smart Signals — find it under Smart Signals in your Fingerprint dashboard.
If you’d like to get started with Smart Signals and Suspect Score for your team, contact us for a demo.
