How to Prevent Browser Fingerprinting: What Works

Most privacy advice is about hiding where you are. Browser fingerprinting is about recognising what you are, and it works even when the where is perfectly hidden. A site measures a few dozen properties of your browser and device, combines them into one value, and uses that value to recognise you on the next visit — no cookie stored, no login, nothing you can clear from a settings menu.

That is why fingerprinting quietly defeats a lot of otherwise sensible setups. You can route traffic through a VPN, block every third-party cookie, and browse in a private window, and still hand over a combination of screen size, fonts, graphics driver and timezone that only you have. WebKit’s tracking prevention policy names the problem precisely: fingerprinting is “tracking based on the properties of the user’s behavior and computing environment, without the need for explicit client-side storage.”

This guide covers what is actually being measured, the two rival defence strategies (they are genuinely opposed, and mixing them makes things worse), what each mainstream browser does today, and the popular “fixes” that increase your exposure instead of reducing it.

What a Fingerprint Is Made Of

The EFF’s Cover Your Tracks project documents the categories it collects, and that list is a fair map of the commercially useful surface:

• The user agent string and HTTP Accept headers — browser, version, platform.
• Screen resolution, colour depth, and window dimensions.
• Timezone and language.
• Installed fonts — historically one of the richest signals, because font sets differ by OS, locale, and installed software.
• Canvas and WebGL hashes — the browser is asked to draw something, and the exact pixels returned vary with GPU, driver, and font rasterisation.
• Plugins, extensions, and media devices exposed to the page.
• Touchscreen support, Do Not Track status, cookie and “super cookie” behaviour.

No single item identifies anyone. Nearly everyone shares a timezone with millions of people. The identification comes from the combination: each attribute narrows the pool, and a handful of moderately unusual values multiplied together is often enough to leave a pool of one. Cover Your Tracks scores exactly this — how distinctive your particular combination is against a database of other users’ configurations.

Two consequences follow, and they are the reason most intuitions about this are wrong:

1. Rare is bad. A niche Linux distribution with unusual fonts and a high-end GPU is far more identifiable than a stock Windows laptop, even though the Linux user is doing more “privacy work”.
2. Changing IP does not help. The fingerprint travels with the browser. Switching exit nodes, moving to café Wi-Fi, or hiding your IP address without a VPN changes the network path and leaves the fingerprint intact. Understanding the split between network-layer and browser-layer identity is exactly the distinction drawn in VPN vs Tor vs proxy — a VPN is a network tool, and fingerprinting is not a network problem.

Two Opposed Strategies: Blend In, or Lie Consistently

There are only two workable defences, and they pull in opposite directions.

Uniformity means making your browser look identical to every other copy of that browser. Tor Project describes the aim as keeping “users blended together, maximizing privacy for everyone” — it limits “the variety within these attributes to reduce distinctiveness,” for example by restricting font enumeration and standardising window sizes. If a million people report the same values, those values identify nobody.

Randomisation means returning slightly different values to every site, every session, so no stable identifier can form. Brave calls this “farbling”: semi-identifying APIs are perturbed using a per-session, per-site seed, so one site sees internally consistent values while two sites see values that cannot be matched to each other.

Each has a real weakness. Uniformity only works while the crowd is large, which is why it is realistic for Tor Browser and hard for anyone else. Randomisation makes you unmatchable across sites but unusual in a crowd — and Brave’s own write-up acknowledges the tradeoff in the other direction, noting that uniformity-based protections are only effective on popular sites.

The practical rule that falls out of this: pick one strategy and do not customise. Every setting you change in a uniformity browser removes you from the crowd it depends on. That is not a slogan; it is the mechanism.

	Tor Browser	Brave	Firefox (ETP Strict)	Chrome / Safari
Strategy	Uniformity	Randomisation	Blocking and reduction	Reduction, varying
Window size handled	Letterboxing into size buckets	Partly	Partly	No
Canvas / WebGL	Restricted, uniform	Randomised per site per session	Restricted in strict mode	Limited
Font enumeration	Restricted	Randomised	Blocked in newer releases	Limited
Cost to you	Slowest; some sites break	Occasional breakage on Maximum	Occasional breakage	Least friction, least protection

Tor Browser: The Strongest Option, If You Do Not Touch It

Tor Browser is the only mainstream browser whose fingerprinting defence is the primary design goal rather than a feature. The most visible mechanism is letterboxing: rather than reporting your true window dimensions, the browser adds margins so the reported size falls into a predictable bucket shared with other users. The buckets are coarse on purpose — Tor Browser starts with a content window rounded to a multiple of 200px by 100px, which the project describes as putting all users into a couple of buckets to make it harder to single them out. Tor’s documentation is equally explicit that making every user genuinely identical is impractical, so the goal is to shrink the number of distinct values that exist at all.

The failure mode is entirely user-driven. Resizing the window, installing an extension, changing a preference, or enabling a font all move you out of the bucket. The Tor Browser safe setup guide covers the operational side of this in detail, including the security levels and the reason the maximised-window habit is a genuine deanonymisation risk rather than a nitpick.

Firefox: Two Different Features, Frequently Confused

Firefox ships two separate things, and treating them as one is a common mistake.

Enhanced Tracking Protection in Strict mode is the supported path. Mozilla’s rollout of expanded fingerprinting protections blocks access to graphics card details, font information, processor core counts, touchscreen capability, and dock or taskbar dimensions. Mozilla states the defences “cut the percentage of users seen as unique by almost half,” and that they debuted in Private Browsing and ETP Strict rather than being on for everyone. Mozilla is also open about the design constraint: timezone stays available because calendar and conferencing tools legitimately need it.

privacy.resistFingerprinting is the about:config flag that ports Tor Browser’s uniformity approach into Firefox. It is stronger and it breaks more, and Mozilla’s own guidance points ordinary users at the Strict-mode protections instead. It is a reasonable choice for a dedicated, single-purpose profile; it is a poor choice for the browser you pay bills in.

Brave, Safari, and Chrome

Brave applies farbling by default across canvas, WebGL, audio, device enumeration, plugins, and the user agent, with three levels: off, default (subtle randomisation that preserves underlying values), and maximum (pure randomisation, with breakage accepted). Default is the sensible setting for daily use.

Safari commits, in WebKit’s published policy, to preventing all covert tracking including fingerprinting, and where prevention would cause undue user harm, to limiting the capability instead — reducing available entropy or restricting time windows. Mitigations are applied universally rather than by exception.

Chrome is the weakest of the four for this specific threat, and the honest framing is structural rather than technical: fingerprinting resistance reduces the identifying signal available to advertisers, which is a conflict of interest for an advertising company. If fingerprinting is in your threat model, Chrome is not the browser to solve it in.

What Does Not Work

These come up constantly and are worth naming, because each one is popular and each one is counterproductive.

• VPNs. They change your IP address. A fingerprint is computed inside the browser and is unchanged by the tunnel. A VPN is still worth having for what it does do — see the ProtonVPN and Mullvad comparison — but it is not a fingerprinting defence.
• Private or incognito windows. They discard cookies and history at the end of the session. Your fonts, GPU, and screen size are identical in a private window.
• User-agent spoofers. Announcing “Safari on macOS” from a browser that renders like Firefox on Linux creates a mismatch that is itself a strong signal. Inconsistent lies are more identifiable than the truth.
• Stacking privacy extensions. Installed extensions are frequently detectable from page context, and an unusual combination of them is a near-perfect identifier. One well-chosen content blocker is better than six.
• Cosmetic customisation of a uniformity browser. Custom fonts, a resized window, and a changed theme in Tor Browser each subtract from the anonymity set the browser is built on.

Test Your Own Exposure

Two public tools measure this directly, both free and neither requiring an account:

• Cover Your Tracks (EFF) reports whether your browser has a unique fingerprint and shows the per-attribute breakdown, which is the useful part — it tells you which signal is making you rare.
• AmIUnique maintains a comparable dataset with a different population, which is a helpful second opinion.

Interpret the result carefully. A “unique” verdict on a small research dataset does not mean a commercial tracker has singled you out, and a “not unique” verdict in Tor Browser is the whole point rather than a fluke. Run the same test in each browser profile you actually use; the differences between your own profiles are usually more instructive than the absolute score.

Match the Effort to the Adversary

Fingerprinting defence has a real cost — broken sites, slower browsing, and constant discipline about not customising things. Whether that cost is worth paying depends entirely on who you are trying to be private from, which is the same question behind every decision on this site. If your adversary is an ad network building a profile, Firefox in Strict mode or Brave at default settings removes most of the commercial value with almost no friction. If your adversary is a capable one that can correlate sessions deliberately, uniformity in Tor Browser is the only strategy that holds, and it only holds if you leave it alone.

Work that out before choosing a browser, not after: start from what a threat model is, or run the interactive threat-model builder to turn your actual adversaries and constraints into a prioritised plan in your browser.

And keep the browser layer in proportion. A perfect fingerprint score does not survive logging into a real-name account, and it does not survive the document metadata problem covered in removing metadata before sharing files. The recurring pattern in OPSEC mistakes that deanonymise people is that the technical layer was fine and a human link gave it away.

Bottom Line

Fingerprinting is identification without storage, so nothing you can delete will remove it. Choose one strategy and commit to it: Tor Browser if you need to be indistinguishable, Brave or Firefox Strict if you need to be untrackable across sites with a browser you can still live in. Change as little as possible afterwards, drop the spoofing extensions, and measure the result with Cover Your Tracks rather than assuming. For most people, one deliberate browser choice removes more tracking than every other setting on this page combined.