← Insights
Note 006 6 min read 1,297 words

Draw the CMMC boundary before you write the SSP

Most of the cost of CMMC Level 2 is decided by scope. Here's how I map where CUI actually goes before any control work starts.

cmmcnist-800-171compliance

Small defense companies usually start CMMC the same way. Someone downloads NIST SP 800-171 Revision 2, opens a spreadsheet with 110 rows, and starts marking requirements as met or not met.

That is the wrong first step. The 110 requirements are fixed. What you control is how many systems they apply to, and that number drives almost every cost that follows: licenses, hardening hours, logging volume, the length of the assessment, and how much of your workforce has to change how they work. Scope comes first. The system security plan describes the scope; it can’t decide it for you.

Start with the data, not the network

Network diagrams show what’s connected. They don’t show where Controlled Unclassified Information actually goes. Before I look at a single firewall rule, I trace the information itself.

For each contract that carries DFARS 252.204-7012, I want answers to these questions:

  • How does CUI arrive? Email from the prime, a download from a customer portal, a shared drive the prime hosts, drawings on removable media, a contract deliverable package.
  • Where does it land first? An inbox, a downloads folder, an engineering workstation, a PLM system.
  • Who works on it, and with what? CAD tools, a document management system, a build server, a test bench.
  • Where does it go next? Back to the customer, to a subcontractor, to a machine shop, to an archive.

Write those answers down as flows, one line per hop. You’ll find paths nobody planned. The classic one is a program manager who forwards marked drawings to a personal phone so they can read them on the road. Another is the shared engineering drive that started as CUI-only and gradually filled up with everything else.

If you’re not sure whether a given document is CUI, don’t guess. The marking requirements come from DoD Instruction 5200.48, and the contract decides what’s covered. That’s a question for your contracting officer or the prime’s security office, and it’s worth asking in writing.

Sort every asset into a category

The CMMC program rule, 32 CFR Part 170, defines asset categories for Level 2 scoping. Each one gets treated differently during an assessment, so the category you assign is an engineering decision with real cost attached.

CUI Assets process, store or transmit CUI. These are assessed against all Level 2 requirements.

Security Protection Assets provide security functions to the CUI environment, whether or not they touch CUI themselves. Your identity provider, the EDR console, the log server, the vulnerability scanner and the remote tools your managed service provider uses all fall here. Teams forget these, and an assessor won’t.

Contractor Risk Managed Assets can process CUI but aren’t intended to, because policy and technical controls keep it off them. You document them in the asset inventory, the SSP and the network diagram. If the documentation doesn’t hold up, the assessor can check further.

Specialized Assets include government furnished equipment, operational technology, IoT devices, restricted information systems and test equipment. They’re documented and managed under risk-based policies rather than assessed against every requirement.

Out-of-Scope Assets can’t process, store or transmit CUI and don’t provide security functions to assets that do. They need real separation from the CUI environment.

The phrase that matters most in that last category is “can’t.” An assessor will not accept “we don’t put CUI on those laptops” as a reason. They want to see the control that stops it: a separate tenant, a blocked sync client, a conditional access rule, a segmented network. If you can’t name the control, the asset isn’t out of scope.

Pick a scoping pattern on purpose

Most small companies end up in one of a few shapes.

Whole company in scope. Every laptop, every account, every cloud service. This works when nearly everyone touches CUI, as in a 15-person engineering shop that does nothing but defense work. It’s simple to explain and expensive to run.

An enclave. A separate environment for the people and systems that handle CUI, with the rest of the company outside it. Less to harden, but the boundary has to be real, and people’s daily workflow changes. The usual failure is CUI escaping through the corporate email system because the enclave didn’t include a way to receive it.

A cloud enclave. The enclave lives in a government cloud tenant, often a virtual desktop that users reach from corporate machines. DFARS 252.204-7012 requires a cloud service provider that stores, processes or transmits CUI for you to meet requirements equivalent to the FedRAMP Moderate baseline, so check the provider’s authorization before you design around it. The endpoint users connect from still matters. Decide how that endpoint is categorized, write down why, and be ready to show what stops copy-paste, file transfer or a local download from moving data out of the session.

No pattern is correct in general. The right one is the smallest boundary your workflow can live with, held in place by controls you can demonstrate.

Test the boundary like an attacker would

Once the boundary is drawn, I try to break it. That’s partly habit from offensive work, and partly because it’s cheap compared to finding the gap during an assessment.

  1. Search for CUI outside the boundary. Run a content search across mailboxes, file shares and cloud storage for the CUI banner and the distribution statements your contracts use. Every hit outside the enclave is either a data spill to handle under your incident procedures or a flow you missed.
  2. Try to move data out. From inside the enclave, attempt to email a test file to an outside address, sync it to a personal cloud drive, copy it to USB and print it. Record which attempts the controls stop.
  3. List everything that reaches in. Admin accounts, MSP tools, backup agents, patching systems, monitoring. Each one is a Security Protection Asset or a hole in the boundary.
  4. Follow the logs. If enclave logs go to a SIEM in the corporate environment, that SIEM is in scope.

Fix what you find, then update the diagram. The diagram, the inventory and the SSP should describe the same boundary. When they disagree, assessors notice quickly.

Write the SSP last

With the flows, categories and test results in hand, the system security plan almost writes itself. Requirement 3.12.4 asks for a plan that describes the system boundary, the environment of operation, how requirements are implemented and how systems connect. You now have real answers for each of those, backed by evidence you generated yourself.

That’s a very different document from one written from a template before anyone looked at where the data goes.

Where this lands in the standards

Scoping for Level 2 is defined in 32 CFR Part 170, and DoD publishes a CMMC Scoping Guide for Level 2 that walks through the asset categories in more detail. The requirements themselves come from NIST SP 800-171 Revision 2, which is the version CMMC Level 2 assesses against, and NIST SP 800-171A describes the assessment objectives an assessor uses for each one. DFARS 252.204-7012 sets the safeguarding obligations that make the CUI flows matter in the first place.

Which CMMC level a contract requires, and when, is set by the solicitation and the contract. That’s a question for your contracting officer or counsel, not your engineer.

Get a second look before you commit

Changing a boundary after you’ve hardened it is expensive. If you want an engineer to check your CUI flows and asset categories before you build around them, start with the CMMC Level 2 readiness gap review. It’s a fixed-scope, requirement-by-requirement look at where you stand against NIST SP 800-171 r2, and scope is the first thing I test. Not sure it fits? Book a free 30-minute fit call.

End of note 006 · 6 of 6