Trained on 55,000 behavioral sessions

Stop friendly fraud
before authorization.

Sentinel scores every checkout on how it was performed, not just what was entered. Keystroke dynamics, mouse motion, form hesitation, and device biometrics — fused into a single risk score in under 40 ms.

92.3%
Accuracy
81.2%
Precision
< 40ms
Inference
Live scoring
TX 918230 · $214.90
High risk
87
model p = 0.912
rule score = 74
Suspicious paste action on card number
IP / billing country mismatch (US → RO)
Abnormal keystroke speed (+42% vs. baseline)
Unusual device touch angle
Behavioral signal fabric

Seven signals. One verdict.

Keystroke dynamics

Flight time, dwell time, and typing rhythm deviation vs. the user baseline.

Mouse motion

Trajectory smoothness, jitter, and speed vs. the account's historical profile.

Form hesitation

Pauses per field, backtracking, and card-number field time.

Touch biometrics

Grip angle, pressure, and stroke geometry on mobile devices.

Copy-paste intent

Detects pasted card data — the strongest predictor of stolen credentials.

IP / billing geo

Country mismatch across IP, billing address, and merchant of record.

Account age & velocity

Session cadence and device-account pairing history.

Behavioral match

Fused similarity to the enrolled biometric template for the user.

01
Capture

A 4KB browser SDK records typing, mouse, and touch signals during checkout.

02
Score

POST /api/v1/verify-behavior returns risk score, status, and flagged reasons in <40ms.

03
Decide

Approve, step-up, or block. Every decision streams live into your dashboard.