A comprehensive study of how fingerprintable users are

A large-scale study of browser fingerprinting uses telemetry from nearly 25 million Firefox users to measure how identifiable browsers are and which characteristics contribute most. Privacy-preserving collection methods, including OHTTP, were used to prevent individual users from being identified or tracked. The analysis finds that uniqueness usually comes from a small number of high-impact measurements, not from one feature alone or dozens of features accumulating: for one modeled service, 80% of uniquely identifiable users could be distinguished using just eight metrics, despite the service collecting more than 50. Results also differ substantially by operating system, so a single overall uniqueness rate can conceal important variation. Shannon entropy alone is not a reliable way to assess a measurement’s fingerprinting impact; the size and distribution of groups sharing fingerprints matter too.

The researchers used these findings to design protections that shipped in Firefox 143, reducing the uniquely identifiable share from 75% to 46% while preserving web compatibility. Their evaluation shows that protections can have sharply different effects across platforms: limiting reported screen resolution, for example, substantially reduced uniqueness on macOS but had little effect elsewhere. Standardizing browser outputs helps users blend into larger groups, but is stronger when paired with randomization; neither approach is sufficient alone. The work also links measurement differences to underlying hardware and user characteristics, enabling more targeted protections. The accompanying site provides the paper, detailed data and code, and an interactive browser test that compares local results with aggregate measurements without transmitting them.