PaleMoon browser has a canvas.poisondata config setting (also available in the Preferences GUI) which prevents such fingerprinting. However, note that these things are not used in isolation - browser fingerprinting is done by collecting various other data too, and when collated together, these can provide enough uniqueness to identify (or categorise) your browser into some specific group. We are fighting a losing battle for privacy.
back
1 comments
Hot take: instead of making fingerprints more unique, we should be striving to make fingerprints as identical as possible, so they stop being useful as fingerprints. In the case of Canvas 2D, that would mean standardized bit-level definitions of what all the canvas operations do, so all the browsers do the same thing.
Unfortunately this could only be done by forbidding web content from using GPUs at all. GPUs are extremely diverse compared to CPUs. It is not feasible to standardize bit-exact operations on diverse GPU hardware while maintaining reasonable performance.
Some will say that we should forbid web content from using GPUs. But that would push apps and their users to native platforms that don't even try to minimize fingerprinting at a technical level, instead either relying on legal enforcement by monopoly gatekeepers (with all the problems that implies), or providing no protection against fingerprinting at all.
> Some will say that we should forbid web content from using GPUs. But that would push apps and their users to native platforms that don't even try to minimize fingerprinting at a technical level
Conflating "websites" with "apps" is what lead to this mess in the first place. You can't just "try to minimize fingerprinting", you either prevent it or don't. And by the time you expose a GPU to the web, you're well inside the latter category.
IMO the only solution to the tracking epidemic is making boundaries between the two clear. Just like some random blog can't read my GPS location without asking first, it shouldn't be given access to other tracking vectors without user consent either.
> You can't just "try to minimize fingerprinting"
Plainly false. You can minimize the number of bits of entropy in the fingerprint even in situations where a couple of bits are unavoidable, and you can mitigate fingerprinting methods by detection and/or blocking. Browsers do this today.
The web is crucial as the only free platform for distributing software to a huge chunk of consumer devices. Apple would love to strengthen their iOS app distribution monopoly by forbidding sophisticated web apps. That's why they have dragged their feet implementing more advanced web standards and limited their capabilities when they do implement them (for example making fullscreen mode unusable for games).
> Plainly false. You can minimize the number of bits of entropy in the
fingerprint even in situations where a couple of bits are unavoidable,
A single API may just yield a couple of bits, but it adds up when there are hundreds of APIs, with new ones introduced every week. And you don't need that many bits to uniquely identify someone.
But sure, leaking a few bits here and there might as well be unavoidable when two of the three major browser vendors are ad companies and preventing it isn't a priority. (See the saga about Google and 3rd-party cookies.)
> and you can mitigate fingerprinting methods by detection and/or blocking. Browsers do this today.
You can mitigate a finite set of fingerprinting methods that you know of. It becomes exponentially harder with every new tracking vector that is enabled by default, especially when the expectation is that things Just Work.
(For example, blocking canvas readout breaks canvas-based image resizing on lots of websites that use the first result from stackoverflow.)
> The web is crucial as the only free platform for distributing software to a huge chunk of consumer devices. Apple would love to strengthen their iOS app distribution monopoly by forbidding sophisticated web apps. That's why they have dragged their feet implementing more advanced web standards and limited their capabilities when they do implement them (for example making fullscreen mode unusable for games).
Respectfully, I don't see a pressing need to solve the issue of "you don't own Apple devices you pay for" by stuffing every possible API under the sun into the browser.
Besides, I'm not advocating against sophisticated web apps; I just wish browsers applied the principle of least privilege when adding features ripe for abuse. e.g. maybe I would allow GPU access for a web-based 3D game whose developer I trust, but not some random blog that will use it to either fingerprint me or run a cryptominer.
> See the saga about Google and 3rd-party cookies
This is a pet peeve of mine. I haven't seen a sane take on this anywhere. Getting rid of 3rd party cookies to prevent tracking has been a priority for Google for many years. Everyone thinks they haven't done it because they hate privacy or something; nothing could be further from the truth. They have been blocked on disabling 3rd party cookies because of antitrust concerns coming from other ad companies who object to being blocked from tracking users.
Antitrust concerns which wouldn't have had any weight if not for yet
another tracking mechanism that Google had intended to add in place of
third party cookies.[0]
It's not because Google "hates privacy", it's because Google operates to generate profit, and it does so from targeted advertising.
[0]: https://www.eff.org/deeplinks/2021/03/googles-floc-terrible-...
As COVID has taught us "if only we had all obeyed the stay at home order" (which would have eliminated COVID and a chunk of dozens or hundreds of infectious diseases), relying on the bulk of man to "get with the plan" is a fool's errand, ipso facto the proposition of such is and was foolish for reason of blindness to this fact.
That is to say, if we all have a "same browser" then there will be people with "not same", there will be divergences, it will be a mess.
A very simple solution of polymorphic fingerprinting - the fingerprinters will get a fingerprint, a good looking one, just that it will be different each time. Even better to think of when the russians poisoned the ammunition in vietnams with bad bullets. Can't make them all bed or they'll toss the whole batch. I think they arrived at 1:10. The idea is to make fingerprinting as ugly as possible - up to and no further than the point that they are forced to cook up something even more evil.
Hotter take: we should provide random answers to all of these measures.
My user agent and canvas size and gpu capabilities should be unique on every single request.
(Within reason, of course)