oss-sec mailing list archives

Re: Fuzzing findings (and maybe CVE requests) - Image/GraphicsMagick, elfutils, GIMP, gdk-pixbuf, file, ndisasm, less


From: Robert Święcki <robert () swiecki net>
Date: Mon, 17 Nov 2014 16:17:33 +0100

2014-11-17 14:10 GMT+01:00 Hanno Böck <hanno () hboeck de>:
Am Mon, 17 Nov 2014 13:48:39 +0100
schrieb Raphael Geissert <geissert () debian org>:

c) fuzz all the tools in there and report at least the
low-hanging-fruit-bugs? (and then maybe try to replace the
"they-don't-fix-bugs-or-don't-have-a-dev-any-more"-tools with more
secure ones)

d) acknowledge the fact that most tools were not "designed for
security" and that we should talk about mitigation. It's about risk
analysis.

Fair point, however it doesn't exclude doing c) as well. (however it
was my understanding that all widely deployed memory corruption
mitigation methods are mostly incomplete and can usually be circumvented
by tricky enough exploits - but I'm no expert on this)

We'll probably never (in a foreseeable future) get to a point in which
gimp/vlc etc. will be safe (to a reasonable degree) to use with
arbitrary files just by the means of fuzzing (or for that matter, by
manual code review, which with such large code-base seems just
unfeasible). The main point of fuzzing seems to be to minimize
existing attack surface before applying decent mitigation techniques
(syscall sandboxes, fault isolation containers etc.)

The reasons for that are:

a). it's a moving target - new code for new and existing media types
is being added every month
b). even if we start fuzzing it now, it seems to me like a temporary
effort (but I can accept that I'm wrong on that)
c). there are too many "this-type-of-tools" to be taken care of with
sufficient attention (file archivers, media players, image converters,
hundreds of file parsers, ...)
d). fuzzing is able to discover only subset of bugs

So, even if after such fuzzing round we ended up with such tools in a
better condition, ergo fewer users will get pwned (which is definitely
a good thing), this by no means should be enough to just use stock vlc
with untrusted media files (if one cares about security of her/his OS)
w/o good evaluation of the risks involved.

I know that this sounds awfully impractical (at least for the time
being, because the landscape here is changing pretty rapidly), but
some would say that the best advice they can give to "average users"
now is to watch "untrusted" movies with web browsers which are
employing well-reviewed and tested sandboxing technologies and their
media decoders are well tested (also: fuzzed). I guess "regular" media
players will follow with this approach in some time.

By no means I want to discourage the effort, just trying discuss its
goals (less avg users pwnd - vs - it's safe to use vlc with everything
now).

-- 
Robert Święcki


Current thread: