What is Amazon Fraud Detector?
Amazon Fraud Detector is a fully managed machine learning service for detecting fraudulent online activities. The service uses the same technology that Amazon.com uses for fraud detection. You train models with your historical data and receive real-time risk scores for new transactions.
Core Features
- Automatic Model Training: ML models are trained on your data without ML expertise
- Real-time Predictions: Millisecond response times for live transactions
- Rule-based Decisions: Combine ML scores with business rules
- Event Types: Specialized models for registration, login, and transactions
- Model Explainability: Understand which factors contribute to fraud scores
Typical Use Cases
Payment Fraud: Assess fraud risk at checkout in real-time. Block suspicious transactions or request additional authentication.
Account Takeover: Detect suspicious logins based on device, location, and behavior patterns. Protect customer accounts from takeover.
Fake Accounts: Identify fraudulent registrations at signup. Prevent abuse of promotions and new customer offers.
Benefits
- Fast implementation without ML team
- Scales automatically with transaction volume
- Continuous improvement through feedback loop
- Reduces fraud while minimizing false positives
Integration with innFactory
As an AWS Reseller, innFactory supports you with Amazon Fraud Detector: data preparation, model training, integration into e-commerce systems, and optimization of decision logic for your risk profile.
Typical Use Cases
Frequently Asked Questions
Do I need my own fraud data for training?
Yes, Fraud Detector requires historical transaction data with labels (Fraud/Legitimate). Alternatively, you can start with Amazon-owned models and add your own data later.
How fast are the predictions?
Fraud Detector delivers predictions in milliseconds for real-time decisions during checkout, login, or registration.
What are Detector versions?
Detector versions enable A/B testing of different models. You can deploy new versions in parallel and gradually redirect traffic.
How does Fraud Detector differ from SageMaker?
Fraud Detector is a specialized service for fraud detection with pre-trained models. SageMaker offers more flexibility for custom ML but requires more expertise.