actionType code. Every vector has specific userData requirements and contributes different risk scores based on telemetry signals, device fingerprint, IP intelligence, graph insights, and deterministic rules.
How It Works
Each request must declare anactionType from the closed set of 17 fraud vector codes. The engine validates the payload, enriches it with device/IP/graph data, evaluates risk factors, and returns a score (0—100) with an APPROVE, REVIEW, or BLOCK decision.
Payment & Financial Fraud
Carding, BNPL, chargeback, promo abuse, seller fraud, money mule, COD abuse, return/refund abuse, loyalty points fraud
Account & Session Fraud
Device farm, bot farming, session hijack, account takeover, password reset abuse, OTP brute force
Content & Referral Fraud
Review bombing, content spam, affiliate/referral fraud
Cross-Cutting Checks
Device integrity, fraud rings, IP anonymization, blocklists — applied to every request

