Now that we’ve talked you through building your own business reference model and building a data criticality matrix, we’re now going to tie it all together and give you the one true secret of information assurance: matching it all up with a security control baseline.

Now the way that NIST has things set up, it works like this:

• Determine the data types (BRM and SP 800-60)
• Determine the criticality for each data type (SP 800-60 and FIPS-199)
• Selecting a security control baseline (SP 800-53)
• Tailoring like a madman

We just did the first 2 steps. Now on to select a control baseline. In the NIST world, it’s easy: pick the right appendix (low, moderate, and high security control baselines) from SP 800-53 and get busy. The part that you don’t see is where NIST took all the applicable regulations and mushed them together into one solid catalog of controls to make it easy for you: do you want vanilla, chocolate, or rocky road? They added in internal controls (ala SoX), regulatory controls from laws like the Privacy Act of 1974, and controls from similar frameworks like BS7799. I should probably say “best practices” in there somewhere but I don’t believe in saying that phrase so I won’t. =)

But we’re now guerillas slaving away in the jungles of information assurance, we cut corners where we need to, and take more time than usual when we need to. We’ve gone fast and furious so far, but this is one step where we don’t need to go fast because what we do next directly determines how much money we spend. Because now we’re going to match up security needs and control requirements into something buildable. We’re going to do some regulatory c*mpliance tracing and control tailoring.

Now have a look at the data criticality matrix we made. See where I started to identify where the various data locations are? This is important, let’s take a look at these columns. Take a system, say, the Knowledge Management System. It’s Column Y, and it can be either a single server, a group of servers, or a whole DMZ set up exclusively for knowledge management–it doesn’t matter as long as we in the end tie it back into pieces of hardware and/or software. It doesn’t have client or incident data on it, which is a good thing. It does have HR data on it (really, what were we thinking when we put it there?) so the overall categorization of the system is MML and the system itself has to c*mply with SoX, Breach Laws, and our internal information security policy. That’s nice, we can deal with that–it’s better than doing Breach Laws for every IT system we have, no matter how big or small.

Need more examples? How about Column X, the Trouble Ticket System. It has trace amounts of client mission data and security incident data which happen to end up there–usually from the help desk who put it in there when a user calls in with a problem or when a tool captures the content of a TCP/IP packet when it triggers on an event like an IDS signature match. Since this is trace amounts of data, what I do is as the client if they are concerned about it. Usually they’re not–the data is not in a high enough aggregate to warrant any kind of controls above and beyond what I do for all my systems–anything else just isn’t that cost-effective to worry over.

One more trick of the trade: identify the governance drivers that apply to all the systems on your matrix. You’ve now just identified the common controls that need to be built for every system no matter what.

Now for each IT system/asset that we’ve identified (and the set of common controls), we can make a list of all the controls that that system has to have, then tailor those to engineering requirements that we can build. I have the start of an example in the CISO Book of Death under the tab “System 0001 Control Matrix”. Basically you put the control objective, where it comes from, and how you intend to build it. It’s exactly like a requirements traceability matrix only with some security words thrown in there. One pet peeve of mine: don’t call security controls requirements, you just confuse and frustrate the people who have to build them.

Now think about it for a minute. Why do we do all this? Well, somewhere in there is savings on licenses, hardware, redundancy, and audits because we just learned how to scope security controls. Less is always better, and what we’ve really done is to tell us where we need more security and where we need less.

Take a step back and guess what? We’ve now just reverse-engineered the beginning to the FISMA approach to information security:

• Break the enterprise down into bite-size pieces
• Determine criticality for each piece
• Determine security control requirements for each piece
• Determine something buildable and testable for each control requirement via tailoring and further definition
• Test to see if all the controls have been built
• Plan to fix the controls that don’t exist or are broken

But you all knew this was where we were headed with this business, didn’t you? The naysayers can now comment on how FISMA and Certification and Accreditation sucks. =)

I also think that maybe I need to spend some more time fishing because I’ve spent way too much time thinking about the BRM and data types.  Sounds like a plan for this weekend.

• And in This Corner, Special Publication 800-60
• In Search of a Better Catalog of Controls
• New SP 800-60 is Out, Categorize Yerselves Mo Better
• Blow-By-Blow on S.773–The Cybersecurity Act of 2009–Part 1
• Some Random Thoughts on C&A SOPs