It's possible to steal data from sessionstore.js including cookies. I'm filing this bug to attach a testcase since bug 413250 is public.
Created attachment 298422 [details] testcase This requires Download Statusbar: https://addons.mozilla.org/en-US/firefox/addon/26
Yea - we should fix this asap...
I worried about that one :-(
trunk no longer uses sessionstore.js, but if it's an upgraded Firefox 2 profile we don't clean up and there may still be juicy stuff available. I'll file a bug on that aspect. I thought we were going to make the Object and Array prototypes immutable to prevent this kind of attack? Is that something that has to wait for JS2?
filed bug 413689 about stale sessionstore.js data in a FF3 profile.
> I thought we were going to make the Object and Array prototypes immutable to > prevent this kind of attack? I thought so too, but I'm still waiting for an answer from Brendan in bug 376957.
Jesse: it's not a mystery. Waldo patched SpiderMonkey to use the original values of Object and Array (and their prototype objects, which while mutable are bound by immutable 'prototype' properties in the ctor objects) when interpreting object and array initialisers. So no replacing or shadowing attacks are possible, but getters on the prototype can still be abused, maybe. /be
Last I remember, shaver's patch in bug 322889 would make sets and gets not use getters/setters -- might be worth seeing what happens with that patch in place, if it's ready enough.
Of course, that's only Array and not Object, so that might not be meaningful here.
Fix for bug 413250 checked in.
http://www.hiredhacker.com/2008/01/19/firefox-chrome-url-handling-directory-traversal/ has just been updated with a demo covering this very facet of the bug: http://www.hiredhacker.com/steal_sessionstore.html Still keeping this private? When can 18.104.22.168 be released?
Release candidate builds are being generated now (see the usual nightly release directories). I'd like to give Window a chance to update http://blog.mozilla.com/security before opening up this bug -- drawing more attention to the problem just puts more people at risk, we're already going as fast as we can on the release.
though chrome: traversal seems fixed on trunk and latest-2.0, resource: traversal still works. resource:///%2e%2e%2f%2e%2e%2f%2e%2e%2f%2e%2e%2f ^^^ this is "/" resource:///%2e%2e%2f%2e%2e%2f%2e%2e%2f%2e%2e%2fetc ^^^ this is "/etc" (in case you don't get /, add more %2e%2e%2f even more: <script src="resource:///%2e%2e%2f%2e%2e%2f%2e%2e%2f%2e%2e%2fetc"> works from the web. this means it is possible to read at least some js files if the full path is known via 'resource:///'.
resource stuff is public Bug 394075 this is none of my business, but i don't see any point in releasing .12 with resource: wide open - it is too similar to chrome: traversal
I've verified this fix with Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:22.214.171.124) Gecko/2008012820 Firefox/126.96.36.199 and Mozilla/5.0 (Macintosh; U; Intel Mac OS X; en-US; rv:188.8.131.52) Gecko/2008012822 Firefox/184.108.40.206. With this testcase (attached to bug), we get "undefined" as a result now. I've also tested and verified bug 413250.