Modeling IT Attacks

Every day IT managers have to contend with an ever-changing risk environment. That's where good risk modeling can help.Late Friday afternoon, EE Times ran a story about ongoing research at NIST and George Mason University that aims to create "attack graphs" designed to help measure the risk posture of an organization: "By analyzing and assigning probabilities to every path a hacker could use to penetrate a computer system, NIST hopes attack graphs will help IT managers identify weak points that need to be patched to safeguard valuable data.

Attack graphs, developed by NIST jointly with George Mason University, calculate the vulnerability of each path into a computer system using NIST's National Vulnerability Database (NVD). By assigning a probable risk to various computer network pathways, the researchers hope to secure computer systems from multistep attacks.

"

Here's how NIST described the technology in its announcement:

"NIST researchers evaluate each route and assign it a risk based on how challenging it is to the hacker. The paths are determined using a technique called "attack graphs." A new analysis technique based on attack graphs was jointly developed by Singhal and research colleagues at George Mason University. A patent is pending on the technique."

This type of threat modeling can go a long way in helping security teams determine which patches they need to deploy first. For instance, by evaluating the vulnerabilities in a network through a scan, you could see that there are many low, medium, and high vulnerabilities scattered about. Fortunately, the low-risk vulnerabilities are on the perimeter, and the highly critical flaws are deep inside the corporate network on a number of servers.

The bad news is that the business managers and application owners don't want those servers patched for another three weeks, because it's so costly and risky to mess with them. What do you do now to mitigate those critical vulnerabilities?

Well, a close look at your attack graph (as NIST is calling it) reveals that by patching a few low-risk vulnerabilities on the perimeter, all risky external access to those at-risk servers has been closed. You can sleep now knowing that those servers are relatively safe until the business will let you apply patches.

Threat modeling like this certainly isn't new, and security vendors such as Skybox Security and RedSeal Systems have been providing modeling capabilities for some time.