Amazon Fraud Detector launches new ML model for online transaction fraud detection
Amazon Fraud Detector is excited to announce the Transaction Fraud Insights model, a low-latency fraud detection machine learning (ML) model specifically designed to detect online card-not-present transaction fraud. Like other Amazon Fraud Detector models, Transaction Fraud Insights leverages more than 20 years of fraud detection expertise from Amazon and AWS. The new Transaction Fraud Insights model type detects up to 30% more fraudulent transactions and maintains its performance up to six times longer than Amazon Fraud Detector’s previous model type, Online Fraud Insights.
Amazon Fraud Detector is excited to announce the Transaction Fraud Insights model, a low-latency fraud detection machine learning (ML) model specifically designed to detect online card-not-present transaction fraud. Like other Amazon Fraud Detector models, Transaction Fraud Insights leverages more than 20 years of fraud detection expertise from Amazon and AWS. The new Transaction Fraud Insights model type detects up to 30% more fraudulent transactions and maintains its performance up to six times longer than Amazon Fraud Detector’s previous model type, Online Fraud Insights.
Online transaction fraud is on the rise. As more merchants transition from brick-and-mortar to online, bad actors are following suit using increasingly sophisticated attacks. Merchants eventually bear the cost of fraudulent charges in the form of chargeback fees, non-refundable transaction fees, lost merchandise, and operational costs. The Transaction Fraud Insights model detects more fraudulent transactions by automatically computing risk patterns such as whether a buyer is making a repeat purchase and how frequently the buyer makes purchases. Amazon Fraud Detector handles the heavy lifting of calculating these values related to a buyer’s history, keeping them updated, and ensuring they are available for fraud predictions and model re-trainings. Transaction Fraud Insights models maintain their performance longer than previous Amazon Fraud Detector models because these auto-computed values are updated in near real-time and used in each low-latency fraud prediction.
To get started, define your transaction event and create your event dataset by uploading your historical transactions using Amazon Fraud Detector’s batch import feature. Next, train your Transaction Fraud Insights model in a few clicks. Once your model is trained, you can embed the Fraud Detector API into your checkout flow to generate fraud predictions in real-time, or use Fraud Detector’s batch prediction API to perform offline or scheduled predictions. For each transaction, Amazon Fraud Detector automatically updates your stored event dataset, which means you can start a model retraining in seconds.
The Transaction Fraud Insights model type is available in the US East (N. Virginia), US East (Ohio), US West (Oregon), Europe (Ireland), Asia Pacific (Singapore) and Asia Pacific (Sydney) regions. For additional details, see our documentation and pricing page.