Fraud Scoring

Fraud Scoring uses machine learning models to assign a risk score (typically 0-100) to each transaction in real-time, predicting the probability that the transaction is fraudulent.

Models analyze hundreds of signals including device fingerprint, IP geolocation, transaction velocity, behavioral biometrics, and historical patterns. Merchants set threshold rules (e.g., block if score > 85, review if score > 60) to balance fraud prevention against false positive rates.