Date: Mon, 7 Nov 2005 11:21:20 -0500 From: Eric Gauthier <eric@roxanne.org> Cc: nanog@merit.edu

I'm not exactly "in the know" on this one, but the heap-overflow advisory that we've seen indicates that the IOS updates Cisco put out are not patches for this problem:

"Cisco has devised counter-measures by implementing extra checks to enforce the proper integrity of system timers. This extra validation should reduce the possibility of heap-based overflow attack vectors achieving remote code execution." from http://www.cisco.com/warp/public/707/cisco-sa-20051102-timers.shtml

We've asked Cisco for a better explanation - namely, are their recommended updates "patches" to the problem (i.e. repairs) or simply mitigating updates that make is harder to exploit. The wording of their advisory seems to indicate the latter. This latter case is what worries me since it implies that there is a fundamental problem in IOS, the problem still exists even after patching, and that Cisco can't readily repair it. Unfortunately, so far we've gotten the run-around and haven't been able to get a better answer, again leading me to believe the worst.

Most exploits (be it IOS or some other target) require multiple things to occur before the "desired effect" is achieved. "buffer overflow" exploits. in general. involve a minimum of two things: 1) "smashing" memory outside of the area you 'should' have been limited to. 2) having 'some other code' accept/use that 'improperly modified' memory believing it to be 'valid' content. Causing =any= step of the exploit process to fail means that the attempt _does_not_succeed_. Re-coding to eliminate all 'possible' buffer overflow situations is a *big* job. The required field-length checking for every multi-byte copy/move operation does have a significant negative impact on performance, as well. Merely _identifying_ the 'tainted' (by being in contact -- directly or in- directly -- with 'user-supplied' data) data-structures is a task measured in man-years. As is isolating _all_ the points where such tainting occurs. Then, and only then, can you begin to -plan- how to remove the taint, whether by sanity-based bounds-checking, 'clipping' to known limits, explicit length checks, or whatever else is appropriate. *AFTER* all that, you can =start= implementing the code that removes taint. It _can_ be much quicker (in terms of "time to delivery to the field") to go after one of the 'other things' that has to happen for an exploit to "work".