[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Tue Apr 12 19:26:45 2005
From: Valdis.Kletnieks at vt.edu (Valdis.Kletnieks@...edu)
Subject: linux bugs (survival stories)?
On Tue, 12 Apr 2005 12:06:59 +0545, Bipin Gautam said:
> every to verify its types) On the other hand we have had products like
> stack guard, lib safe, grsecurity.net, se-linux
> BUT i was woundering, to what extent adding these extra security
> measures are effective against the real attacks & bugs discovered in
> the kernel.
They do almost nothing to guard against bugs discovered *in the kernel*,
because all of them are addressing *userspace* bugs.
There's pretty much *zero* you can do to protect against kernel bugs, other
than audit the code and get rid of the bugs. Remember it's basically a
monolithic kernel - that means that if a bug lets somebody get control of
the kernel, it's Game Over.
There's a reason why binary kernel modules are frowned upon - and it's precisely
that. As far as the kernel is concerned, a security bug and a buggy binary module
are the same thing - both can go scribbling on whatever kernel data structures
it wants to, and there's no real way to stop it.
If anybody wants a good kernel-auditing project, just start going through the
2.6.12-rc2 tree and look at uses of copy_from_user(), and make sure that each
use of that function then proceeds to *validate* the data (especially in the
various driver's .ioctl methods - historically a place for issues). At least
the tree seems to be a lot better with not ignoring the return value of
copy_from_user() ;)
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 226 bytes
Desc: not available
Url : http://lists.grok.org.uk/pipermail/full-disclosure/attachments/20050412/64982c40/attachment.bin
Powered by blists - more mailing lists