Directly Asking the Security Data
We have long been fans of network forensics tools to provide a deeper and more granular ability to analyze what’s happening on the network. But most of these network forensics tools are still beyond the reach (in terms of both resources and expertise) of mass markets at this point. Rocky D of Visible Risk tackles the question, “I’m collecting packets, so what now?” in his Getting Started with Network Forensics Tools post.
With these tools we can now ask questions directly of the data and not be limited to or rely on pre-defined questions that are based on an inference of subsets of data. The blinders are off. To us, the tools themselves aren’t the value proposition – the data itself and the innovation in analytical techniques is the real benefit to the organization.
It always gets back to the security data. Because any filtered and/or normalized view of the data (or metadata, as the case may be) is inherently limited because it’s hard to go back and ask the question(s) you didn’t know to ask at the beginning of the investigation, query, etc. When investigating a security issue, you often don’t know what to ask ahead of time. But that pretty much breaks the model of SIEM (and most security, by the way) because you need to define the patterns you are looking for. Of course we know attackers are unpredictable by nature, so it is getting harder and harder to isolate attacks based on what we know attacks look like.
When used properly, network forensic tools can fundamentally change your security organization from the broken alert-driven model into a more effective data-driven analytic model.
It’s hard not to agree with this position, but the details remain squishy. Conceptually we buy this analytics-centric view of the world, where you pump a bunch of security data through a magic machine that finds patterns you didn’t know where there – the challenge is to interpret what those patterns really mean in the context of your problem. And that’s not something that will be automated any time soon, if ever.
But unless you have the data the whole discussion is moot anyway. So start collecting packets now, and figure out what to do with them later.