Ref: https://learn.cantrill.io/courses/1820301/lectures/42194142

• Summary:
• Contents:

Amazon Fraud Detector - Key Concepts

• 🔧 Fully-managed fraud detection service
  • e.g. new account creations, payments, guest checkouts
• Model types:
  1. Online Fraud (little historical data)
     • e.g. new customer account (anything suspicious around this particular sign-up?)
  2. Transaction Fraud (uses transactional history)
     • e.g. identify suspect payments
  3. Account Takeover
     • e.g. phishing or other social based attacks
• Things are scored → Rules/decision logic allow you to react to a score based on business activity or business risk

Amazon Fraud Detector - Example Architecture

Ref: https://www.udemy.com/course/aws-certified-machine-learning-engineer-associate-mla-c01/learn/lecture/45320483

1. Use API to ingest data from S3
   • Can perform some ETL and feature engineering at this stage
2. Select a ML model
3. Fraud Detector automatically trains, tunes, validates, and deploys model
4. Use API to infer data from deployed model
5. Can continuously learn over time as new data comes in
6. Can provide insights into importance of features (model variables)