Amazon Fraud Detector

How Amazon Fraud Detector works

Amazon Fraud Detector analyzes our data using classification machine learning. We provide our historical fraud data, and Amazon Fraud Detector uses this data to train, test, and deploy a custom fraud detection model.

Let’s look at the key components of Amazon Fraud Detector that analyze our data, learn from it, and make accurate predictions about potential fraudulent activities.

• Event: An event is an organization’s business activity that we want to evaluate for fraud risk.

• Fraud model: A fraud model results from machine learning algorithms analyzing the provided event data. These algorithms learn patterns indicative of fraudulent behavior and use them to make predictions.

• Model Training: Model training uses a provided event dataset to create a model that can predict fraudulent events.

• Model Deployment: Model deployment is a process for activating a model version and making it available for generating fraud predictions.

Amazon Fraud Detector can be used in various scenarios, such as detecting fraudulent online payments, identifying new account fraud, preventing trial and loyalty program abuse, and improving account takeover detection. Let's discuss one use case in detail below:

Use case: Detecting new account fraud

One of the key use cases of Amazon Fraud Detector is detecting new account fraud. This involves distinguishing between legitimate and high-risk account registrations to selectively introduce additional checks such as phone or email verification.

In the digital age, it’s common for scammers to create fake accounts to exploit online services. They may use these accounts to carry out illegal activities, abuse promotional offers, or even compromise the security of other users.

Amazon Fraud Detector helps mitigate this risk by analyzing the data associated with new account registrations. This data can include the user’s IP address, email, device information, and more. The service uses machine learning models trained on historical fraud data to identify patterns and behaviors associated with fraudulent accounts.

For example, Amazon Fraud Detector can flag these as high-risk registrations if many accounts are created from the same IP address or if the email addresses follow a suspicious pattern. Activities with a high-risk score are routed for human verification, while those with low-risk scores are promptly sent to the client application. The service can then trigger additional verification steps for these accounts, such as sending a verification code to the user’s email or phone.

This helps prevent fraudulent activities and enhances the security of our online services. It ensures that only legitimate users can access our services, thereby improving the overall user experience.