← All articles

Fingerprint Entropy: Why Being Rare Is the Problem

2026-08-25 · 6 min read

Entropy measures how much a signal narrows the crowd around you. A value shared by half of all visitors removes one bit; a value held by one visitor in a thousand removes about ten. Being rare is what makes you identifiable, which is why every attempt to look unusual works against you.

What is a bit of entropy, concretely?

One bit halves the population. If half the visitors to a site run Windows, learning that you run Windows removes one bit and leaves you inside half the crowd. Learning your exact screen resolution might remove four or five bits. Signals stack, and about 33 bits is enough to single out one person among eight billion.

Thirty-three bits is the whole reason fingerprinting works without cookies.

Signal	Rough entropy	Effect on the crowd
Operating system family	~1–2 bits	Halves it, roughly
Browser and major version	~2–4 bits	Narrows to a version cohort
Screen resolution	~4–5 bits	Narrows to a display class
Timezone	~3–4 bits	Narrows to a region
Installed font list	~8–15 bits	Often nearly unique
Canvas rendering	~8–10 bits	Narrows to a hardware and driver class

Why does hiding harder make it worse?

Because rarity is the currency. Blocking a signal, spoofing an unusual value or installing a heavy privacy stack all produce a configuration few other people have, and a rare configuration is a high-entropy one. You have not removed information; you have replaced a common value with a distinctive one.

Rarity as a cost is the counterintuitive centre of the whole subject, and it catches almost everyone at least once.

Which values are actually low-entropy?

The boring ones. A recent stable browser version, a 1920×1080 screen, a default font set for the operating system, a timezone matching the IP and an ordinary consumer GPU. These are shared by millions of people, so each one narrows the crowd very little and none of them stands out.

Choice	Entropy contributed	Outcome
Common resolution, common OS	Low	Blends in
Rare resolution such as 1337×917	High	Stands out
Default font set	Low	Blends in
200 designer fonts installed	Very high	Nearly unique
Canvas returning a plausible hash	Moderate	Normal
Canvas blocked or throwing errors	High	Flagged as unusual

How does entropy relate to account separation?

Two ways at once. Each profile needs low entropy so it does not stand out on its own, and the profiles need enough distance from each other that they do not read as one device. Twenty profiles that all look like the same boring machine are as linkable as twenty profiles that all look strange.

The target is a set of plausible, ordinary, mutually different configurations — which is harder than either extreme and is the actual work a profile tool does.

Does a bigger crowd always help?

Only against fingerprinting, and not against everything else. Blending into millions of similar devices defeats identification by rarity, and it does nothing about cookies, a logged-in session or an IP address that has carried 5 other accounts.

Treat entropy as 1 of 3 layers rather than the whole problem. A perfectly ordinary device on a burned address linked to a shared payment method is still connected — entropy work fixes the layer it fixes, and the other 2 need their own attention.

How do you check where you stand?

Run a fingerprint test that reports per-signal rarity rather than only a hash. The useful output is not "your fingerprint is unique" but which specific values are doing the narrowing, because those are the ones worth changing.

In practice the biggest contributors are usually fonts and canvas, and both are fixable — trim an unusual font collection back toward the platform default, and make sure canvas returns something plausible rather than being blocked. Two changes tend to move more than a dozen smaller ones. Check once at profile creation and once after each browser update, then stop.

Entropy is a setup problem rather than an ongoing one, and repeatedly adjusting a profile that already blends in is how a stable device becomes an unstable 1.