1852541 - meet.google.com - Cannot join meetings when Canvas Randomization is enabled

Status: VERIFIED FIXED

(Core :: Privacy: Anti-Tracking, defect, P2)

Description

[Raul Bucata]

From github: https://github.com/webcompat/web-bugs/issues/126901.

<!-- @browser: Firefox 119.0 -->
<!-- @ua_header: Mozilla/5.0 (Windows NT 10.0; rv:109.0) Gecko/20100101 Firefox/119.0 -->
<!-- @reported_with: unknown -->
<!-- @public_url: https://github.com/webcompat/web-bugs/issues/126901 -->

URL: https://meet.google.com

Browser / Version: Firefox 119.0
Operating System: Windows 10
Tested Another Browser: Yes Safari

Problem type: Something else
Description: Cannot join meetings with privacy.resistFingerprinting set to true
Steps to Reproduce:
Unable to attend any Google meet meetings with privacy.resistFingerprinting set to true not reproduced on Chrome or Safari.
<details>
<summary>View the screenshot</summary>
<img alt="Screenshot" src="https://webcompat.com/uploads/2023/9/6ebb44fb-690b-4fef-8a1e-3cc8de9c4536.jpeg">
</details>

<details>
<summary>Browser Configuration</summary>
<ul>
<li>None</li>
</ul>
</details>

From webcompat.com with ❤️

Change performed by the Move to Bugzilla add-on.

Comment 1

[Raul Bucata]

The issue was reported via the webcompat.com reporter. Since the issue is reproducible with an altered pref, privacy.resistFingerprinting set to TRUE, we have moved the issue accordingly. Please feel free to move the issue to the relevant Project and Component.

Comment 2

[David Hu]

Hello, I was the original reporter. I also reproduced this issue with privacy.resistFingerprinting set to false but privacy.fingerprintingProtection set to true.

Comment 3

[Harshit Sohaney [:hsohaney]]

Can we close this bug since this pref isn't supported?

Comment 4

[Tim Huang[:timhuang]]

I think we should still leave this bug open for tracking the issue. Ideally, we should find the root cause and then decide on the following action to do here.

Comment 5

[Tom Ritter [:tjr]]

FPP is supported. I can reproduce it with FPP enabled, and override testing reproduced it with -AllTargets,+CanvasRandomization. (For -AllTargets it did not reproduce.) Will try to dig into why though....

Comment 6

[Tom Ritter [:tjr]]

Spent a while debugging this, unsuccessfully from the JS code side of things, so I switched to using modified Firefox and looking specifically at the canvas calls from that side. Meet was rendering two canvases: one 400x200 that made a very typical fingerprinting image, and one 5x5 with a constant pattern. Trial and error showed that if we didn't modify the 5x5 image, then Meet would work. My conclusion is that they are checking for randomization or some other error.

I did experience inconsistency with my testing. Sometimes the page failed even though no pixels were modified. Sometimes it passed with only 1 or 2 pixels modified.

Comment 7

[Tom Ritter [:tjr]]

Also, I want to thank Raul and everyone for bringing this bug to our attention, it would have been a big issue to leave unaddressed.

Comment 8

[Tom Ritter [:tjr]]

Attachment: Bug 1852541: Do not randomize the canvas if it is one uniform pattern r?timhuang

Comment 9

[Tom Ritter [:tjr]]

Attachment: Bug 1852541: Fix tests and add one for constant canvas r?timhuang

Depends on D188133

Comment 10

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/autoland/rev/6dff3776d3d5
Do not randomize the canvas if it is one uniform pattern r=timhuang,jgilbert
https://hg.mozilla.org/integration/autoland/rev/dbb5ce808936
Fix tests and add one for constant canvas r=timhuang

Comment 11

[Cosmin Sabou [:CosminS]]

Backed out for causing build bustages on nsRFPService.cpp.

Push with failures: https://treeherder.mozilla.org/jobs?repo=autoland&resultStatus=testfailed%2Cbusted%2Cexception%2Cretry%2Cusercancel&revision=dbb5ce808936ca231a5ed941c6ebbc083596d01d&selectedTaskRun=Z44aPgurRY6533bGq8-Ivw.0

Failure log: https://treeherder.mozilla.org/logviewer?job_id=430257460&repo=autoland

Backout link: https://hg.mozilla.org/integration/autoland/rev/cf0ddbec5963b58732cc2ce17a8064ea31d45fb1

Comment 12

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/autoland/rev/9dea05c869be
Do not randomize the canvas if it is one uniform pattern r=timhuang,jgilbert
https://hg.mozilla.org/integration/autoland/rev/7f516f079dd4
Fix tests and add one for constant canvas r=timhuang

Comment 13

[Iulian Moraru]

Backed out for causing build bustages + bc failures.

• Push with failures

• Failure log for build bustages

• Failure log for bc failures example 1

• Failure log for bc failures example 2
  Also there were this TV failures:

• Failure log for TV job

• Backout link

Comment 14

[Tom Ritter [:tjr]]

Attachment: Bug 1852541: Improve the odds of Canvas Randomization bit-flipping r?timhuang

Our current algorithm gives a lot of test failures and test-verfiy failures.
After some logging and debugging, my best guess is that we are just getting
unlucky.

I would see test failures (failed to detect a randomization was applied) when
we e.g. fliped a mere 16 bytes from 0 -> 1. When we flipped some values from
0 -> 2 that would more reliably cause a change.

Comment 15

[Tom Ritter [:tjr]]

Attachment: Bug 1852541: Clarify some function names r?timhuang

While debugging I felt these would be more descriptive function names

Depends on D190283

Comment 16

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/autoland/rev/8832d7e6c0dc
Do not randomize the canvas if it is one uniform pattern r=timhuang,jgilbert
https://hg.mozilla.org/integration/autoland/rev/4e369c26c1f6
Fix tests and add one for constant canvas r=timhuang
https://hg.mozilla.org/integration/autoland/rev/21d022246ce0
Improve the odds of Canvas Randomization bit-flipping r=timhuang

Comment 17

[Cristian Tuns]

Backed out for causing build bustages in nsRFPService.cpp

• Backout link
• Push with failures
• Failure Log
• Failure line: /builds/worker/checkouts/gecko/toolkit/components/resistfingerprinting/nsRFPService.cpp:1373:24: error: code will never be executed [-Werror,-Wunreachable-code]

Comment 18

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/autoland/rev/0af1e383e35e
Do not randomize the canvas if it is one uniform pattern r=timhuang,jgilbert
https://hg.mozilla.org/integration/autoland/rev/187a1f13809f
Fix tests and add one for constant canvas r=timhuang
https://hg.mozilla.org/integration/autoland/rev/eae33625ba1e
Improve the odds of Canvas Randomization bit-flipping r=timhuang

Comment 19

[Narcis Beleuzu [:NarcisB]]

https://hg.mozilla.org/mozilla-central/rev/0af1e383e35e
https://hg.mozilla.org/mozilla-central/rev/187a1f13809f
https://hg.mozilla.org/mozilla-central/rev/eae33625ba1e

Comment 20

[BugBot [:suhaib / :marco/ :calixte]]

The patch landed in nightly and beta is affected.
:tjr, is this bug important enough to require an uplift?

• If yes, please nominate the patch for beta approval.
• If no, please set status-firefox119 to wontfix.

For more information, please visit BugBot documentation.

Comment 21

[Tom Ritter [:tjr]]

This doesn't affect Beta (unless someone has modified preferences) because Canvas Randomization is only enabled in Nightly.

Comment 22

[Simona Badau, Desktop QA]

Verified as fixed on the latest Nightly 120.0a1 - I'm not seeing any errors when joining Google Meets calls while the preference "privacy.fingerprintingProtection" is enabled - tested on macOS 11.7, Ubuntu 20.04 and Windows 10.

Comment 23

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/autoland/rev/35e20e70da91
Clarify some function names r=timhuang

Comment 24

[Narcis Beleuzu [:NarcisB]]

https://hg.mozilla.org/mozilla-central/rev/35e20e70da91