Fraud Detector

Amazon Fraud Detector was a managed service that trained a fraud-detection model on a customer’s own event history and scored transactions or sign-ups against it.

It is closed to new customers. AWS stopped accepting new customers on 7 November 2025. Existing customers can continue to use the service; a new AWS account cannot adopt it.

What to use instead

AWS points fraud-detection workloads in two directions depending on the problem.

Transaction and behavioural fraud — train a model on your own data with AutoGluon, the open-source AutoML library, and deploy it on Amazon SageMaker AI. AutoGluon handles the tabular model selection and ensembling that Fraud Detector previously did behind its API, and SageMaker AI hosts the resulting model behind an endpoint. This costs more design work than Fraud Detector did but leaves the model, its features and its thresholds under your control.

Account-creation and login abuse — AWS WAF Fraud Control protects sign-up and sign-in pages against credential stuffing, credential cracking and fake account creation. It works from managed rules rather than a machine learning model, so it addresses a narrower problem than Fraud Detector did, but it needs no training data and sits in front of the application rather than inside it.

If you are already a customer

Existing event data can be exported; AWS documents a migration path for event data stored in Fraud Detector. Plan the move rather than waiting, since a service closed to new customers receives no new features.