Post

Large-scale measurement of browser fingerprinting in the wild

A study using telemetry from nearly 25 million Firefox users models how a small set of browser characteristics can make people identifiable. The researchers report that protections introduced in Firefox 143 reduced the modeled unique share from 75% to 46% for one commercial fingerprinter.

Those are population-level estimates, not universal tracking probabilities: results vary significantly by operating system and fingerprinting method. The team also measures the size of groups sharing a fingerprint, since merely counting unique users misses how small the remaining groups can be.

HN readers compared results from the interactive test and discussed unusual browser configurations and GPU-based fingerprinting.