[18991] in cryptography@c2.net mail archive
Re: "ISAKMP" flaws?
daemon@ATHENA.MIT.EDU (Peter Gutmann)
Sat Nov 19 20:51:33 2005
X-Original-To: cryptography@metzdowd.com
X-Original-To: cryptography@metzdowd.com
From: pgut001@cs.auckland.ac.nz (Peter Gutmann)
To: paul.hoffman@vpnc.org, smb@cs.columbia.edu
Cc: cryptography@metzdowd.com
In-Reply-To: <20051118151809.CDEBD3BFE60@berkshire.machshav.com>
Date: Sat, 19 Nov 2005 13:44:44 +1300
"Steven M. Bellovin" <smb@cs.columbia.edu> writes:
>In message <p062309a8bfa266327e1b@[10.20.30.249]>, Paul Hoffman writes:
>>Which "proper programming tools" would check for a logic path failure
>>when a crafted packet includes Subpacket A that is only supposed to
>>be there when Subpacket B is there, but the packet doesn't include
>>Subpacket B? There are no programming tools that check for this, or
>>for related issues: it has to be the implementer who has enough
>>understanding of the protocol and enough time (and program space) to
>>code against such issues.
>
>Decent test case generators.
The problem is that these are extraordinarily labour-intensive to write.
Admittedly they're incredibly effective in finding problems (every time
someone's gone to the effort of creating one, it seems like 90% of all
implementations in the target area have proven vulnerable), but that still
leaves the problem of creating the things in the first place.
Another issue is that all of the current ones (that I know of) test for random
rather than Byzantine failures, i.e. they create large numbers of random
packets and hope that one of them triggers a bug, rather than carefully
crafting malicious payloads designed to cause faults. Once we get Byzantine
test-case generators, I predict there'll be another round of security alerts
as 90% of the products out there fail yet again.
Peter.
---------------------------------------------------------------------
The Cryptography Mailing List
Unsubscribe by sending "unsubscribe cryptography" to majordomo@metzdowd.com
