AWS in the race for fraud

The AWS Fraud Detector tool announced at the Re: Invent 2019 conference is now available. Based on a machine learning engine, it can detect fraud in online payments and the creation of fake customer accounts.

Online fraud is a real nuisance. The market for solutions in this area, valued at more than $ 19 billion in 2018 by research firm Fortune Business Insights, is expected to grow 25.5% annually through 2026 and reach $ 110 billion by that date. . In this market, players like IBM, DXC Technologies, and Experian are well established, but we must now expect a significant new addition: AWS.

The Fraud Detector tool that was mentioned during the last Re: Invent 2019 event last December has now been announced in general availability. The latter enables the fight against online payment fraud and the search for fake customer accounts. “Organizations doing business online are particularly vulnerable to attack by criminals with a variety of modes of operation including creating fake accounts and stealing credit card payments,” said AWS. ÔÇťAmazon Fraud Detector uses machine learning (ML) and leverages over 20 years of fraud detection experience from AWS and to automatically identify questionable cyber practices so you can spot more fraud. fast. With Amazon Fraud Detector, you can build a fraud detection model with just a few clicks and no previous ML experience, as Amazon Fraud Detector does most of the ML work for you. “

Fraud detection in ML increased or not

“We recently started using Amazon Fraud Detector and we are pleased that it offers a cost-effective implementation and a self-service approach to creating a machine learning model that is personalized for our company,” said a feedback . by John Kercheval, Senior Director of the Identity Services Group at GoDaddy, in a video. “The model we built in Amazon Fraud Detector can potentially detect fraudulent registrations instantly. So we’re very pleased with these results and see what else we can do.” Amazon didn’t say if he was a user of his own solution himself, but we can very strongly assume that.

Fraud Detector is based on pre-built ML models to identify fraud related to online activities such as online payments, creating fake accounts, and ordering guests. A fraud detection console allows the user to create rules (acceptance, rating, etc.) and flag suspicious accounts for in-depth analysis. A number of APIs that can be integrated with third party applications are also available as it is known that a fraud detection model can be imported into Fraud Detector from Amazon SageMaker.

In terms of pricing, AWS charges its tool by the hour ($ 0.39 for model training and $ 0.06 for hosting) and $ 0.03 per prediction for real-time fraud information. online (400,000 first predictions per month) with a sliding scale afterwards. Without the use of machine learning, the price of predicting fraud based on simple rules is not surprisingly lower, starting at $ 0.005 for the first 400,000 predictions per month.