1485249 - WebGL extensions should be disabled when private.resistFingerprinting is enabled

Status: NEW

(Core :: Graphics: CanvasWebGL, enhancement, P2)

Description

[Arthur Edelstein [:arthur]]

In Tor Browser, to protect against fingerprinting we disabled WebGL extensions by setting the pref "webgl.disable-extensions" to true.

We would like to propose that when "privacy.resistFingerprinting" is true, WebGL extensions should be disabled as well, regardless of the other pref.

Comment 1

[Arthur Edelstein [:arthur]]

We also set "webgl.disable-fail-if-major-performance-caveat" to true. We could apply this behavior when "privacy.resistFingerprinting" is true as well.

Comment 2

[Tom Ritter [:tjr] (OOTO until 4/30 at least)]

Attachment: Bug 1485249 - Let ResistFingerprinting also apply the functionality of two WebGL prefs r?jgilbert

Comment 3

[Tim Huang[:timhuang]]

Unassign myself because I am no longer actively working on this.

Comment 4

[Marcos Dione]

Either that, or have a toggle button to enable it (because it should be disabled by default; according to EFF's panopticlick, WebGL is the biggest leaker of info).

Comment 5

[Marcos Dione]

How much effort do you think this should take for someone entirely not familiar with FFox's source code?

Comment 6

[Marcos Dione]

I just disable webgl, and panopticlick can still do some webgl ops to the point of producing a fingerprint. Should I open a new bug about this?

Comment 7

[Simon Mainey]

(In reply to Marcos Dione from comment #6)

I just disable webgl, and panopticlick can still do some webgl ops to the point of producing a fingerprint. Should I open a new bug about this?

13ae805231fcd00154a46b5a992143ec is their way of saying webgl is disabled: it's hash of nothing. That it currently says 1 in 9 indicates that (not that you should read too much into entropy figures here: the data sets are not real world and are tainted). It also says no javascript for webgl vendor & renderer.

This ticket is about disabling webgl extensions, not webgl itself - that is Bug 1428034

Comment 8

[Simon Mainey]

updated info: tor browser

• ignore: webgl.disable-extensions = true : deprecated FF74+ Bug 1477756
• ignore: webgl.disable-fail-if-major-performance-caveat = true: default true FF86+ Bug 1678652
• webgl.enable-webgl2 = false
• webgl.default-antialias = false is an interesting proposal

Tor Browser has webgl behind their security-slider, and blocks readPixels() regardless. I don't think TB's model fits Firefox. TB wants to reduce entropy in webgl parameters which doesn't matter because webgl is effectively blocked - Firefox can't do that.

Maybe we can converge a little: Firefox RFP to allow a site exception for readPixels(), similar to canvas: default block and limit parameters with webgl2 and antialias

Where are we at with software rendering? e.g. Bug 1724558