"adds random noise to hidden canvas invocations" sounds as clearly signaling you're using something-like-canvas-defender, which is likely very unusual and thus makes you more unique.

The noise is different every hour, meaning that basically the canvas can't be used as a signal. It's not a perfect solution (my useragent string is more unique, which annoys me) but it is better than a 100% identifiable canvas.

There are add ons to randomize your User Agent string. Just set it to something like Windows & Chrome v whatever

> Canvas Defender

My understanding is that these don't actually help. Even their own blog suggests that there is a lot of limitations[0]. And they even talk about how TOR just returns 0. But if everyone did that (say FF did that overnight), then wouldn't that help solve the problem?

Or does there need to be a way to force these things to be pixel perfect (data perfect for the other fingerprinting things like battery, mic, etc)?

[0] https://multilogin.com/blog/

Thank you for this - I'd often assumed Canvas Defender was no good but I couldn't quite put my finger on why. I would encourage anyone with even a passing interest to read this truly illuminating post because there is some fascinating stuff that even as a web developer I did not know:

https://multilogin.com/everything-you-need-to-know-about-can...

It is really fascinating how all this works. And I'm really happy that multilogin writes these blog posts. I think it brings up questions of how necessary is this information and is it worth the cost (loss of privacy). I'm not a web developer so I'm not sure how useful these metrics are. Or is it just limitations? How much of an impact would implementing pixel perfect renders be (as in page load time)? I'd love to know what people that work in these areas think. Pros and cons.

Canvas Defender is Multilogin's product :) Other canvas plugins block the canvas, which is a much clearer signal, and they go to lengths to point out how simply blocking the canvas is worse.

By adding random noise every hour, it basically means that the confidence in the canvas as a signal is going to be 0. That's about the best you can do.

I tend to fail these uniqueness tests pretty catastrophically, and I hadn't heard of Canvas Defender. But solutions like that seem a bit heavyweight to me (and more likely to affect the UX negatively). Is this really the only way to get around this problem?

Yeah, I do too, my useragent string is somehow almost completely unique. I try using a more "common" one and suddenly everything locks me out.

Canvas Defender only adds noise on hidden canvases, ones that are obviously being used for fingerprinting.

But yeah there's no perfect solution. If you simply block canvas rendering, like some plugins do, that is a 100% identifiable signal. So the best you can do is make the fingerprinting "unsure" that that is you, by having the fingerprint be different every time.

> Canvas Defender for Firefox

Anyone aware of something comparable for Safari (on macOS)?