AI service
Amazon Fraud Detector
Builds and deploys models for online fraud detection.
Key points
- Provides managed ML workflows for detecting online fraud such as fake accounts or suspicious transactions.
- Uses event data, entities, variables, labels, rules, and detectors to score fraud risk.
- Combines ML model scores with rules so applications can make fraud decisions.
- Targets online fraud patterns rather than general anomaly detection or network security.
- AWS closed Amazon Fraud Detector to new customers on November 7, 2025.
When to use it
- Choose Fraud Detector in legacy exam-style scenarios about online transaction or account fraud.
- Use it conceptually when event history and fraud labels are available for risk scoring.
- Use SageMaker, AutoGluon, AWS WAF, or other current approaches for new fraud detection architectures.
Exam tips
- Fraud Detector is specific to fraud risk; Personalize recommends items and Forecast predicts time series.
- Rules plus ML model scores are a key service clue.
- Because it is no longer open to new customers, avoid treating it as the default current choice for new builds.
- AWS WAF helps block web attacks, while Fraud Detector focused on ML fraud predictions.