Browsing

Fingerprinting

How websites identify you without cookies, why blocking it is hard, and the strategies that actually reduce it.

On this page
  1. How it works
  2. The counterintuitive part
  3. A note on "fingerprint testers"
  4. Where this fits

Browser fingerprinting identifies you by measuring your browser rather than storing anything in it. Delete every cookie you like. If your browser's combination of measurable properties is unique, you can be recognized on the next visit anyway.

How it works

Scripts can read or infer, among other things: screen resolution and color depth, time zone, language list, installed fonts, GPU model (via WebGL), audio-processing quirks, how your specific hardware renders text onto a canvas element, touch support, and the browser version itself. No single property identifies you. The combination frequently does. The EFF's Cover Your Tracks project demonstrates this live against your own browser and reports how many bits of identifying information each property leaks.[1]

Fingerprinting is invisible: nothing is stored on your machine, there is nothing to delete, and no permission prompt appears. It survives private mode, cookie clearing, and VPNs. A VPN changes your IP, not your GPU.

The counterintuitive part

You cannot fight fingerprinting by customizing your browser to be "more private" one tweak at a time. Each unusual setting, a spoofed user agent, a disabled API, a rare extension leaking its presence, makes your fingerprint more distinctive. Fingerprinting resistance is a crowd property: you are protected when your browser is indistinguishable from many other people's, which is precisely what casual tweaking destroys.

There are two coherent strategies:

Blend into a crowd (strongest). Tor Browser and Mullvad Browser make every user present nearly identical values: standardized window sizes, fonts, time zone handling, and blocked or normalized APIs. Your fingerprint matches thousands of other users, so it identifies nobody. The cost is breakage and inconvenience, a price worth paying when the threat model calls for it.

Reduce and randomize (pragmatic). Firefox's fingerprinting protections and Brave's "farbling" add small randomizations to canvas, audio, and similar readouts so each session measures differently. This defeats a good share of commercial fingerprinting without much breakage. It does not defeat a determined adversary and isn't designed to.

For most people: use a browser with built-in fingerprinting defenses (Brave, Firefox with strict protections), block the tracking scripts outright with uBlock Origin (a fingerprinting script that never loads learns nothing) and resist the urge to install five anti-fingerprinting extensions. For anonymity-grade needs, only the crowd strategy is credible.

A note on "fingerprint testers"

Sites that grade your fingerprint (including Cover Your Tracks) measure uniqueness within their visitor population, which skews heavily toward privacy enthusiasts. A "unique" score there is informative, not a verdict. Use them to understand which properties leak, not as a pass/fail exam.

Where this fits

Fingerprinting is one linkage channel among several. If you log in to a site, the fingerprint is moot, you identified yourself. The realistic goals are: deny cheap cross-site tracking (achievable with the pragmatic strategy), or deny linkage between separated identities (requires the crowd strategy plus the discipline described in Compartmentalization).


  1. Electronic Frontier Foundation, Cover Your Tracks. ↩︎