What a browser fingerprint is, and what PATANYX changes

You ran the test and got a list of readings. Some came back empty. Some came back with your real fingerprint. This page explains what those readings are, why anyone wants them, and what PATANYX does about a few of them.

No single reading identifies you. The combination does.

No single item in a browser fingerprint identifies you. Your screen resolution is shared by millions of people. So is your time zone. So is the list of fonts installed on a stock operating system.

The identification comes from the combination. A tracker collects thirty or forty readings, and the odds that another visitor matches all of them drop fast. The result is a value that stays the same across sites, across sessions, and across a cleared cookie jar, because none of it was ever stored on your machine in the first place. It was measured.

That is what separates fingerprinting from cookies. There is nothing to delete, because nothing was stored.

Where the readings come from

Some readings are simply reported. Screen size, color depth, the number of CPU cores, the amount of memory, the language and time zone settings, the list of available fonts. The browser hands these over because web pages have legitimate reasons to want them.

Others are measured by making your machine do work and watching the result:

• Canvas. The page draws text or shapes into an invisible image and reads the pixels back. Different graphics stacks, drivers and font renderers produce subtly different pixels for the same drawing instructions.
• Audio. The page runs a sound through a processing graph without playing it and reads the numbers that come out. Small differences in how audio is computed show up as a stable value.
• WebGL and graphics hardware. The page can ask for the name of your graphics adapter and driver, and can render a test scene and read it back the way it reads back canvas. PATANYX changes only the two parameters that name your GPU; the rest of WebGL, including direct pixel readback, is deliberately left alone so real WebGL applications keep working.

These are the interesting ones, because they describe the machine rather than the settings. Changing your language or resizing your window does not move them.

What PATANYX changes

For canvas, audio and graphics hardware, PATANYX gives every site a different reading.

That is noise rather than uniformity, and the distinction is worth a moment. One approach to fingerprinting is to make every user look identical, so that a reading carries no information. PATANYX takes the other approach: every site sees a different fingerprint, so the readings do not join up between sites into a single trail.

It does not cover everything, and it is not a cloak. The explainer page lists exactly what it does not cover. As of the current build, web workers are not covered. For the full list of what is and is not included, read the fingerprint divergence explainer rather than assuming from this page, which deliberately names only one gap.

What the test measured

The test runs the tricks a tracker uses on your own browser and shows you what each one got: nothing, a hint of a protection, or your real fingerprint.

It runs locally and sends nothing. It also runs in any browser, which is the point: you can check our claim against the browser you already use. Run it where you are now, and then run it again somewhere else, and compare the two lists yourself instead of taking a claim about either one on trust.

How to read your results

A reading that returned your real fingerprint is not an alarm. It means that surface is not covered, which for most browsers is most surfaces. A reading that came back as nothing means the page could not measure it at all.

The useful exercise is not scoring your browser out of ten. It is noticing how many separate, boring-looking details a page can collect in the time it takes to load, and that you were never asked about any of them.

If you want to see the complete list of covered and uncovered surfaces, it is on the fingerprint divergence explainer. To run the measurements again, or run them in another browser, use the test page. PATANYX itself is on the download page.