Security: CSS :visited with mix-blend-mode can leak browser history |
|||||||||||||||||
Issue descriptionAn exploit is described in https://lcamtuf.blogspot.com/2016/08/css-mix-blend-mode-is-bad-for-keeping.html which tricks the user into revealing their browser history by clicking on an image whose visibility depends on overlaid <a> elements having a specific combination of :visited states. Repro: http://lcamtuf.coredump.cx/whack/
,
Apr 18 2017
Adding ojan since he's worked on stuff like this before.
,
Apr 18 2017
Given that this was a public disclosure (by a Googler) from last year, any reason to view-restrict this?
,
Apr 19 2017
I agree we don't need to view-restrict this.
,
Apr 19 2017
,
Apr 20 2017
,
Apr 20 2017
Having chatted with our Predictability effort leads there really isn't much we can do at this stage. The usage is too high and the security impact low to warrant a deprecation/removal. All we can do is take this as one other sign that we should rethink how :visited works.
,
Apr 20 2017
I assume you mean the usage of :visited is too high to remove it entirely. But could there be a more targeted mitigation, like using the non-visited background color in mix-blend-mode computations?
,
Apr 20 2017
Sorry no I meant mixed-blend-mode usage here.
,
Apr 20 2017
I'm not too sure what the best way forward is. rbyers@, do you have any thoughts?
,
Apr 20 2017
,
Apr 20 2017
IMHO it's time someone took a serious look at overhauling how :visited behaves to try to solve these issues. It's been discussed a lot but I couldn't find an existing bug so filed bug 713521.
,
Apr 20 2017
,
Apr 20 2017
In that case, what is the action required on this bug (as opposed to bug 713521)?
,
Apr 28 2017
,
May 5 2017
Dropping the update frequency on this to match the blocking bug 713521, since there doesn't seem to be anything for us to do until that is resolved. Also removing the "Privacy" component, since that's captured by the blocking bug.
,
Dec 6 2017
,
Dec 6 2017
,
Dec 6 2017
,
Feb 14 2018
+msramek for Privacy and +some more CSS OWNERS.
,
Feb 14 2018
I certainly vote for issue 713521 as a "once and for all" solution.
,
May 8 2018
|
|||||||||||||||||
►
Sign in to add a comment |
|||||||||||||||||
Comment 1 by nparker@chromium.org
, Apr 18 2017