Backing American Express Customers: How Machine Learning Halts Takeover Fraud

At American Express, our work managing fraud and account takeover risk puts us on the front line of protecting Card Members. We do this by staying one step ahead of fraudsters as they try to create new tricks aimed at deceiving consumers and merchants and leveraging cutting-edge technologies to help stop them and ensure Card Members stay safe.

Global Fraud Risk Decision Science and Global Fraud Risk Management teams lead this effort, and the company has maintained the lowest U.S. fraud rates among major card networks for the past 16 years, according to the February 2023 Nilson Report.

Protecting Amex Card Members

The financial services industry has witnessed a rise in attempted account takeover fraud, where fraudsters use stolen credentials to gain access to a user’s account. Criminals use unconventional methods to circumvent barriers such as chip-and-PIN and one-time online passwords to access card accounts online and commit fraud.

At Amex, we developed a fraud protection system that uses machine learning techniques to identify potential account takeovers with high precision and predict whether online logins are from a genuine customer. While machine learning may sound futuristic, it’s an integral part of our day-to-day efforts to help protect American Express Card Members and involves using algorithms and statistical models to help us understand patterns in data.

Employing Machine Learning

To help stop account takeover fraud, we used various data sources, including new ways to leverage existing data to help identify potential fraudster logins. We then employed state-of-the-art machine learning techniques to arrive at a real time model that scores every single login and predicts fraud risk, almost instantaneously. As a result, we can score the perceived risk of logins and help protect our customer’s accounts. For example, if our model considers an attempt to be high risk, customers need to provide additional proof of authenticity before gaining access to their accounts. Alternatively, the model ranks low-risk logins accordingly and allows customers to gain access to their accounts.

These steps strike a balance between the desire to offer a customer-friendly online experience with the need to protect our American Express Card Members.

This is just one example of the kind of technology we employ as we adapt to changing trends. Our models are robust, and my team is ever vigilant, meaning we’ve got the back of Amex Card Members wherever they are.