Security Watch

Old Attack Exposes Microsoft Shortcomings

One if by Land, two if by code, I on the opposite shore will be ... taking a nap.

• By Russ Cooper
• 03/28/2005

Land attacks first came to light in 1997, and the flaw was fixed in all Windows versions at the time. The vulnerability appears to have been re-introduced as a result of the security hardening done during XP SP2 development.

The fact that the newest versions of Microsoft's OSes can be crashed by Land attacks makes you realize how far Bill Gates' vaunted Trustworthy Computing initiative still has to go. Some key failures this vulnerability exposes:

1. This is an old and well-known form of attack. How could Microsoft miss this during security testing of the new versions? The likely answer is that they had routers which prevented the LAND attacks; thus, they probably tested for the vulnerability and missed the problem because the router blocked the attack, even though the OSes won't. This means the criteria they used for determining the success or failure of the test was completely off-base.
2. Code was originally written, then subsequently found to be vulnerable to an attack (in this case, the original LAND attack in 1997). Such an occurrence should get logged in such a way as to ensure that the issue, or the coding that led to the vulnerability in the first place, was double-checked every time the modules containing the code were revised, updated or replaced. Yet here we are again, so obviously there were no sticky notes on the vulnerable modules saying "Hey, check and make sure we're not vulnerable to LAND attacks!"
3. We could simply attribute this to the age-old charge that Microsoft's code is so huge and so diversely managed that it doesn't know what it's doing with—or to—it.
4. It could also be that its code is just too difficult to do proper quality assurance (QA). It seems hardly fair to blame Computer Associates for legacy issues in its code (as I did a few weeks ago) and not call out Microsoft's QA people for re-introducing a previously patched vulnerability. Holy Windows, Batman!

More information on the Land attack can be found here.

Hacking
Grsecurity, a group of security applications and infrastructure for Linux, has a vulnerability in its Role-Based Access Control (RBAC) sub-system. The vulnerability could allow someone operating as the Root user to gain privileges over other processes, a situation grsecurity is supposed to prevent.

The issue here isn't so much the vulnerability itself; grsecurity provides significant protection against so many forms of attack that this one vulnerability is unlikely to result in a compromise. The real problem is that trust in the grsecurity model is temporarily broken. Systems employing grsecurity are usually highly sensitive, and should be patched immediately to reinstate the trust in the grsecurity environment.

Malicious Code
Instant messenger worms are on the rise. Interestingly, these aren't based on the graphic image vulnerabilities disclosed over the past few months, as one might expect, but instead plain old get-the-user-to-click-on-this type attacks. They're probably gaining traction because they almost always come from someone you know, propagating via buddy lists in your instant messenger program.

Here's a tip. If you get a link from someone via your instant messenger program, simply send them a reply and say "Huh?" If they don't reply immediately, don't click on the link—ever.