Pages

Thursday, July 07, 2005

Rootkit Detection, and a prediction (of sorts)

I was over on Rootkit.com again today, reading up on some of the recent entires...if you're at all interested in Windows security, you should really consider signing up. Anyway, I was reading an article by the erstwhile Joanna Rutkowska on crossview-based rootkit detection, and was really fascinated by what I was reading. Her article on Rootkit.com discusses various issues and does a good job of outlining the "war of attrition" as the good guys develop new ways to detect rootkits, and the rootkit authors (some can be good guys, too...) develop new ways to avoid detection.

Joanna makes some interesting comments, in particular:

One may ask a simple question now: why bother to hide files at all? Isn’t the idea of “hide in the crowed” equally stealth? The answer, fortunately, is no...

and

The answer is no, because the current antivirus technology is able to find all (unhidden) executable files and then perform some kind of analysis if the given PE file looks like a potential rootkit/malware installer (for e.g. check if it uses functions like OpenProcess(), OpenSCManager(), ZwSetSystemInformation() and similar). When designing such scanner we need to remember that rootkit executable can comprises of two parts, one being an actual malware loader and the other being a (polymorphic) decoder.

I'm not sure that I entirely agree with her statement, though her reasoning is certainly sound. First off, let me just say that we all come from different places and have different opinions based on different experiences. For example, I have military training in my background, which includes the concept of Maneuver Warfare (as practiced by experts). Given that, and also given that we're seeing more and more attacks in the media that seem to take a more economic or financial focus, my thought is that we're going to see more targetted attacks.

What does this mean? Well, rather than going for mass infections, we'll likely see programs installed on fewer, but targetted machines. Am I saying that this is the death of Internet worms? Not at all...we'll have those around for a long while yet. But what I am saying is that it's very likely that the worms will be test cases...what works, how "noisy" is it, how quickly is something detected and turned over to the anti-virus vendors for analysis and signature creation? With this kind of information, the attacker can target his approach...and all without rootkit technology.

I'll give you an example. In the military, it's commonly known that during inspections, you give the inspector something to find...b/c if you don't, he won't leave until he's gone to some very dark, uncomfortable places with a microscope and a pen light. So, you give him something to find...not too significant, but enough to satisfy him so he'll...well...go away. Well, map that sort of thing over to what we've been seeing since the inception of viruses, and especially since backdoors like Back Orifice were released...when the incident occurs, it's detected b/c it has a significant and often immediately noticeable impact on systems. Well, what if the attacker decided to be really stealthy, and not give the inspector (Administrator, in this case) any cause to even look around in the first place?

Why are rootkits used? To hide the attacker's presence when the administrator or investigator comes looking. So...don't do anything to cause the investigator to look in the first place.

Where are attacks going? Think about maneuver warfare...one of the concepts is to bypass strongpoints. Marines assaulting a beach will bypass a bunker that's facing the beach, and cut it off from the rear, choking off the supply routes that keep the guys in the bunker in beans, bullets, and band-aids. The same holds true with crime...bad guys are going to attack the easy targets first...the unlocked cars and houses, the unescorted children and women, etc. Online, something that looks like it's fairly unattended/unmanaged will be attacked first. Why go after the heavily protected server, where *if* you do get in, you'll create a lot of noise in doing so (in my book, I used the example of Ethan Hawke in Mission Impossible crushing up a light bulb and spreading the shards outside the apartment in the safe house...), and someone's going to come looking.

Attacks are likely going to be targetting less well protected systems, and the attacks are likely going to have less of an impact on the systems over all. The attacks will be more subtle, and the attacker is going to take great pains to stay stealthy and hidden, by not attracting attention to the fact that he's there. Do you need rootkit technology for this? No. It's been widely seen that it doesn't take a lot of effort to remain hidden from most administrators, even if you're hiding in plain sight (no disrespect intended, guys and gals). Adding a program to a system that isn't going to be detected by anti-virus software (all that takes is something new), isn't going to create a lot of noise, and isn't going crash or overwhelm the system is all it takes.

Are you like me, and need examples and specifics? No problem. Anyone remember

No comments:

Post a Comment