Detector

Source: src/AWS/FraudDetector/Detector.ts

An Amazon Fraud Detector detector — the container that binds an event type to versioned rule sets and models used to evaluate fraud. Creating the detector is cheap; the rules, models, and detector versions that produce predictions are provisioned separately.

Creating a Detector

Basic Detector

const detector = yield* FraudDetector.Detector("checkout", {
  eventTypeName: purchase.name,
});

Detector with an Active Version

const detector = yield* FraudDetector.Detector("checkout", {
  eventTypeName: purchase.name,
});

const version = yield* FraudDetector.DetectorVersion("v1", {
  detectorId: detector.detectorId,
  status: "ACTIVE",
  rules: [
    {
      ruleId: "high_risk",
      expression: '$email == "fraud@example.com"',
      outcomes: [review.name],
    },
  ],
});

Runtime Predictions

Bind GetEventPrediction in the init phase (providing the GetEventPredictionHttp layer on the Function effect) and score events at runtime against the detector’s ACTIVE version.

// init
const getEventPrediction = yield* FraudDetector.GetEventPrediction(detector);

// runtime
const { ruleResults } = yield* getEventPrediction({
  eventId: "order-123",
  eventTypeName: "purchase",
  eventTimestamp: new Date().toISOString(),
  entities: [{ entityType: "customer", entityId: "cust-1" }],
  eventVariables: { email: "buyer@example.com", ip: "1.2.3.4" },
});