Amazon Fraud Detector
Amazon Fraud Detector is a fully managed service that uses machine learning to identify potentially fraudulent activities and accurately distinguish between legitimate and high-risk transactions. It uses your data and the same technology that Amazon uses to protect its own business from fraud.
APIs
Amazon Fraud Detector API
The Amazon Fraud Detector API provides programmatic access to create and manage detectors, models, event types, entities, labels, outcomes, rules, and variables for automated fr...
Capabilities
Amazon Fraud Detector Real-Time Detection
Orchestrate ML models and business rules for real-time transaction fraud scoring and decision-making.
Run with NaftikoFeatures
Automatically trains and deploys ML models using your historical transaction data without requiring ML expertise.
Returns fraud scores within milliseconds for integration into transaction approval flows.
Online Fraud Insights (OFI), Transaction Fraud Insights (TFI), and Account Takeover Insights (ATI) pre-trained model types.
DETECTORPL rule language allows writing conditional logic using model scores and event variables.
Variable importance scores explain which factors most influenced a fraud prediction.
Uses Amazon fraud experience to provide immediate predictions even with limited historical data.
Ingest historical labeled events to continuously improve model accuracy over time.
Use Cases
Score credit card and payment transactions in real-time to block fraudulent purchases.
Detect unauthorized login attempts and account compromise using behavioral signals.
Identify fraudulent new account registrations at signup to prevent synthetic identity fraud.
Flag users abusing discount codes, referral bonuses, and promotional offers.
Reduce chargeback rates by blocking high-risk transactions before they complete.
Score insurance claims for fraudulent patterns in real-time during claim submission.
Integrations
Store training datasets and export labeled event data to S3 for model training.
Control access to detectors, models, and predictions using IAM roles and policies.
Combine Fraud Detector with SageMaker for custom ML pipelines and model integration.
Monitor prediction volumes, model performance, and API error rates.
Encrypt training data and model artifacts with customer-managed KMS keys.
Route fraud detection outcomes to downstream systems for automated response workflows.
Send real-time fraud alert notifications to operations teams via SNS topics.