Skip to main content
Fraud detection is only as good as the data behind it. In AntifraudAI, events are the structured data payloads your application sends to the API — describing what just happened, who did it, and where. Signals are the derived risk indicators that AntifraudAI computes from that event data by combining it with device intelligence, behavioral history, and network context. Together, events and signals form the raw material for every risk score and decision the platform produces.

​
Event Types

AntifraudAI supports three event types, each designed for a distinct category of user action. Transaction events represent financial or payment actions — a card charge, a bank transfer, a crypto withdrawal. Transaction events carry payment-specific fields like amount, currency, payment_method, and billing address information. They are typically screened at the moment of authorization, before funds move. User events capture account-level lifecycle actions: login attempts, new account registrations, password resets, email or address changes, and similar identity actions. Screening user events lets AntifraudAI build a behavioral baseline for each user over time and detect account takeover attempts early. Custom events are fully flexible and let you define any action that matters to your business — promo code redemptions, referral link clicks, API key generation, content publishing, or anything else. Custom events accept arbitrary metadata, so you can include whatever context is relevant to your detection logic.

​
Core Event Fields

The following fields are accepted across event types. Including as many as possible improves scoring accuracy.

​
Signals

Signals are computed automatically by AntifraudAI during event enrichment — you don’t define them, you receive them. Each detected signal is returned in the signals array of the API response and contributes to the event’s risk score. Signals are purely informational in the API response — they tell you why a score is high, but they don’t directly determine the decision. That mapping is handled by the ML models and your configured rules.

​
Enriching Events

Include as many fields as possible in every event payload, especially ip_address and device_fingerprint. These two fields unlock the majority of AntifraudAI’s enrichment capabilities — geolocation, VPN detection, device history, and cross-user velocity checks all depend on them. Events submitted without these fields will produce lower-confidence scores.

​
Metadata

The metadata field accepts any valid JSON object and is your escape hatch for business-specific context that doesn’t fit the standard schema. For example:
Metadata is stored alongside the event and is visible in the dashboard case view. It can also be referenced in custom rule conditions using the metadata.* field path — for example, metadata.account_tier eq premium — giving you fine-grained control without modifying your core event schema.