Cerberus, White Courtesy Phone, Please

Remember what I wrote last time about the danger of assumptions and bias in security monitoring? Well, forget what I said.

No, not really. But there’s another way to look at it. The purposes of monitoring can be many and varied; one of the big ones, of course, is catching the intruder. When most people think of monitoring, they think of this one, and a whole industry has been growing that focuses on deep expertise in this area: attacker methods, indicators of compromise, attribution, and fast response. Take a regular SIEM, layer some catching-the-intruder expertise on top in the form of customized tools, people, and analytics, and you have a more specialized monitoring system.

But monitoring isn’t just about the sexy targeted attacks. If you are the Society for the Preservation of Historic Sites, then you don’t care about the APT, nor does the APT care about you. Why would you still need to monitor?

First off, there’s the automated intrusion, which simply looks for the equivalent of a stick propping the door open. If they get in, and you happen to have something of value to steal or use, they’ll take advantage of that opportunity. No matter how small and non-sexy your organization might be, you should be monitoring for this kind of activity.

There are also general malware attacks, which can rely on phishing, incautious browsing, and good old viruses to work (yes, they’re still out there). Targeted and automated attacks often involve dropping some kind of malware, so there is a whole raftload of companies that specialize in detecting malware (or insulating the system from it). Since this can happen to anybody, you need this kind of monitoring.

What about external attacks that don’t involve malware? We’re getting expertise for those, too, in the form of behavioral analysis and anti-fraud detection. Using a compromised account can look exactly like normal business activity; the only difference is that it’s the wrong person using the account. Conversely, you can have the right person using the account, but he is using it to do the wrong things (such as accessing confidential documents, running his own business off the company PC, or approving and cashing checks for nonexistent purchase orders).

When you start to look at behavior, though, you have to understand a lot more about the application and business layers in order to monitor effectively. Many monitoring systems put together a lot of data to try to figure out whether someone is the wrong person based on network-level or OS-level activity, but for fine-grained detection, you need data from past usage of the application. Has the user ever used the application in this manner? Should anyone be using it this way? Are the requests too close together to be coming from a human’s hands on the keyboard, or are they for pages in an order that doesn’t make sense in the business logic?

This is where even more specialized expertise is popping up in the form of products or add-ons for healthcare, e-commerce, energy, and manufacturing. When you are looking for the "wrong person doing the right things," the "wrong person doing the wrong things," or "the right person doing the wrong things," your system needs to understand more about what "wrong" is.

So sometimes it is important to put bias into your monitoring -- or at least a defined perspective, so that you’re managing the risks that are most pertinent to you. This doesn’t mean you shouldn’t take a step back every now and again to see what you might be missing. Yeah, I never said this stuff was easy.

Wendy Nather is Research Director of the Enterprise Security Practice at the independent analyst firm 451 Research. You can find her on Twitter as @451wendy.