Stripe tightens transparency around fraud-detection data
Fraud prevention is a core part of Stripe’s payment infrastructure. The company has invested in its Radar machine-learning system since 2015, and today it uses a mix of automated models and optional custom rules to block fraudulent transactions worth hundreds of millions of dollars each month. As the online economy expands, so does the sophistication of fraud, which keeps Stripe’s engineering teams working on accuracy and new defenses.
A key input to that defense is a set of real-time signals collected by Stripe.js and the company’s mobile SDKs. These signals—such as screen resolution and browsing patterns—are primarily used to identify bots, i.e., scripted browsers that attempt fraudulent transactions. Because they are processed in real time, these signals measurably improve outcomes: Stripe reports they reduce expected fraud-driven chargebacks by 11%, a significant gain in a domain where small improvements matter.
In response to recent questions about how these signals are gathered and used, Stripe has made several updates to clarify its practices:
- Its privacy and cookie policies now state explicitly that advanced fraud detection signals are used only for security and anti-fraud purposes. Stripe does not—and will not—share or sell this data to advertisers.
- A new documentation page details exactly what advanced fraud detection signals are collected.
- Site and app owners now have the option to disable the collection of these signals altogether.
The changes are intended to make Stripe’s commitment to its users explicit, without altering the underlying technology or its effectiveness at stopping fraud.
