Decoding FIPS 199: A Framework for Categorizing Federal Information Security

Federal information-system categorization begins with a question more precise than “How sensitive is this data?”: What would the loss of confidentiality, integrity or availability actually do to the mission, the people who rely on it and the organization operating the system? A service may contain mostly public information and still be mission-critical if its outputs are changed or it becomes unavailable. Another system may remain operational during an outage but cause serious harm if private records are disclosed. Those are different impact decisions and they should not be collapsed into one vague risk label.

FIPS 199 supplies the common security-category language for federal information and information systems. The practical work is identifying the information types inside an authorized system boundary, assigning reasoned impact values to each security objective, taking the appropriate high-water marks and documenting who approves the result. That category informs subsequent controls and assessment activities; it is not an automatic certification that the system is secure.

Keep three impact objectives separate from the start

The FIPS 199 standard, finalized in February 2004, defines potential impact in terms of adverse consequences if confidentiality, integrity or availability is lost. Its three impact levels are Low for limited adverse effect, Moderate for serious adverse effect and High for severe or catastrophic adverse effect. These are judgments about consequences, not numerical probabilities that an attacker will succeed. A security category therefore does not replace a threat assessment or a vulnerability review.

Confidentiality concerns unauthorized disclosure, including harm to people whose information must be protected. Integrity concerns improper modification or destruction, including unreliable official decisions based on corrupted data. Availability concerns timely and reliable access to information and services. Assess each loss separately: a public incident-status page may tolerate everyone reading it, but not a malicious alteration of evacuation instructions or a prolonged failure during an emergency.

FIPS 199 provides a useful notation: SC information type = {(confidentiality, impact), (integrity, impact), (availability, impact)}. For an information type, confidentiality may be Not Applicable where disclosure genuinely creates no adverse impact. The standard does not permit Not Applicable for integrity or availability. Nor may a whole information system use Not Applicable for any of the three objectives: the system’s processing functions and operational information require at least Low protection even if its user-facing content is public.

That difference is easy to miss when a team tries to give a database a single sensitivity tag. A label such as “public” may describe the release status of one dataset, but it says little about whether unauthorized changes would disrupt agency operations or harm individuals. Document what a failure would mean in the actual mission context rather than treating classification labels as substitutes for analysis.

Inventory information types inside an actual system boundary

A sound categorization needs a defensible system description first. Identify the authorizing boundary, major functions, external exchanges and owners of information the system processes, stores or transmits. The SP 800-60 Volume I, Revision 1 guidance, finalized in August 2008, explains the mapping process from mission and support information types to provisional impact values. It is guidance for choosing and documenting categories, not a promise that every agency has the same mission or should retain the suggested values unchanged.

The companion SP 800-60 Volume II, Revision 1 provides cataloged information-type examples and suggested security categories. Use them as starting points and check the actual information, contractual and legal context. One shared platform may hold public guidance, personnel details, audit events and internal routing or key-management information. Each can have different potential impacts. Data in a vendor-managed service also requires a boundary and ownership decision; a cloud contract alone does not tell the authorizing official which information types or inherited functions belong in the system authorization.

For every identified type, record where it originates, who is accountable for it, which mission or business process it supports, relevant legal obligations, where it crosses the system boundary and why a particular loss would matter. The first proposed category can then be raised or lowered when mission facts support a different conclusion, with the change and approver recorded. Do not select Low simply because a system is small, nor High simply because someone has labeled a field “important.” Consequences and institutional context must support the label.

Consider a fictional civilian service called Eastford Assistance Portal. It publishes public appointment information, stores procurement contact records and retains internal event logs for service accountability. The example is designed to show a calculation; it is not an assessment of an actual federal agency or a real system boundary.

Calculate the system’s three high-water marks, not one average

The essential calculation is in FIPS 199’s information-system rule: for each security objective, assign the highest impact among the information types resident in the system. The result is still a three-part system security category. An average across information types would hide the most consequential loss; making every objective High because one field is High would erase meaningful differences in protection needs.

Illustrative information type in Eastford Confidentiality Integrity Availability Why this might be reasonable
Public appointment notices Not Applicable Moderate Moderate Disclosing already-public notices adds no harm, but altered or missing notices can seriously disrupt access to services
Restricted procurement contacts Moderate High Low Disclosure could harm affected parties; forged supplier instructions could create severe mission or financial consequences
Internal service audit events Low Moderate Low Log tampering can seriously impair accountability even if ordinary access interruptions are limited
Initial system high-water mark Moderate High Moderate Highest impact is taken separately in each column, with system-level minimums and any justified adjustments considered

For this invented set of judgments, the initial category is SC Eastford = {(confidentiality, MODERATE), (integrity, HIGH), (availability, MODERATE)}. The public notices’ Not Applicable confidentiality value does not lower the procurement records’ Moderate confidentiality requirement. The audit events’ Moderate integrity value does not reduce the procurement records’ High integrity impact. No arithmetic averaging or majority vote is involved.

The calculation is only a starting point. Teams must consider whether system-level components introduce a greater impact than the individual information types suggest. The standard itself illustrates increasing a system confidentiality rating to account for sensitive system functions. Likewise, an integration that can submit binding instructions to another agency may require a reassessment of integrity. A reviewer should be able to inspect the cause, the change to the relevant objective, the revised vector and the accountable approval.

The NIST Risk Management Framework Categorize step makes the expected governance outcome explicit: document system characteristics and the completed information/system security categorization, then obtain review and approval by the authorizing official. A spreadsheet that performs a maximum calculation is not, by itself, an approved authorization decision.

Translate the category into an overall impact level without losing detail

FIPS 199 produces the three-objective security category. FIPS 200, finalized in March 2006, uses the highest of those three objective impact values to determine the system’s overall impact level for selecting an initial security-control baseline. A Low-impact system has Low across all three objectives. A Moderate-impact system has at least one Moderate and none High. A High-impact system has at least one High. This is a second high-water mark: across the objectives, after the information-type maxima were calculated.

Eastford is therefore an overall High-impact system in this illustrative example because integrity is High. Its confidentiality and availability values remain Moderate in the recorded vector. Calling the entire vector “High/High/High” would be inaccurate, even though the overall system impact level is High for baseline-selection purposes.

Layer of the decision Recorded output What must not be inferred
Information type Three objective impact values for each distinct type, with justification A type marked public has no integrity or availability needs
Information system Three maxima, with at least Low for every system objective and justified system-level adjustments Every objective must equal the single highest value
Overall system impact Low, Moderate or High from the highest system objective That the system is authorized, safe, or compliant because a label is present
Initial controls Corresponding security baseline, refined for the real system and applicable requirements Baseline selection equals completed implementation or assessment

That distinction matters in review meetings. A system owner should explain which objective drives High, what would happen if it were compromised, and what other objectives still require protection. The overall impact level is useful for consistency, but the vector and its rationale remain essential for planning controls, contingencies and evidence.

Choose and tailor controls only after the impact decision

Once the category and overall impact have been determined, NIST’s SP 800-53B Control Baselines provides Low, Moderate and High security-control baselines, plus a separate privacy baseline. It describes these as starting points for selection and tailoring rather than one-size-fits-all proof of implementation. Its 2025 Release 5.2.0 changed the document version for consistency with related publications without changing the baseline selections. Selecting High does not mean automatically deploying every possible control or proving that every control operates effectively.

The RMF Select step requires teams to select, tailor and document controls, decide which are system-specific, hybrid or common, allocate responsibility and prepare plans for review. For a shared service, the team must identify the controls inherited from the provider and the controls that remain the agency’s responsibility. A cloud vendor’s generic compliance report is not a substitute for checking those inherited assumptions against the actual service boundary.

The underlying SP 800-53 Revision 5 control catalog is a source for specific security and privacy controls, not another impact-categorization formula. The published Release 5.2.0 of August 27, 2025 updated selected controls and discussions, but NIST explicitly reported no SP 800-53B baseline changes. Keep the categorization standard, baseline document and control-catalog revision distinct in the decision log. Do not quietly substitute an old draft or a superseded baseline page when claiming current guidance.

Even after tailoring, authorization requires more work. Implementing and assessing controls, explaining residual risk and obtaining the responsible official’s decision are separate RMF activities. The FIPS 199 category is not a substitute for them. Nor should a general private-sector company infer a federal legal duty merely because FIPS 199 offers a useful analysis method: applicability depends on the system, governing authority, contracts and other requirements. Classified and national-security systems also have specific governing regimes beyond the ordinary scope described by the standard.

Keep the categorization traceable when the system changes

One frequent failure is to treat categorization as an unchangeable property assigned when a system first received a security plan. A new API may begin transmitting sensitive records; a public-status component may become the authoritative source for a safety-critical workflow; an agency may consolidate services that previously had separate authorization boundaries. Each change can affect an information type, a consequence assumption, an objective’s high-water mark or the overall system impact.

The SP 800-37 Revision 2 RMF, finalized in December 2018, places categorization within an ongoing lifecycle of control selection, assessment, authorization and monitoring. It does not prescribe that a particular invented example has already achieved federal compliance. A sound change record links new system facts to an updated impact rationale and explains whether control selection, inherited responsibilities, the security plan or an authorization decision requires further review.

For Eastford, a useful evidence record would contain the approved boundary, an information-type register, three justified impacts per type, the column-by-column maxima, any system-level adjustments, the overall FIPS 200 impact and the version of NIST guidance consulted. Add named information owners, the accountable reviewer, the authorizing official’s decision and review triggers. An impact value without a mission consequence and an approver is not enough for a later reviewer to reconstruct the determination.

Review trigger Question to reopen Evidence to retain
New personal or protected information type Could unauthorized disclosure now cause more harm? Data-flow change, information owner, applicable obligation and revised confidentiality reasoning
New downstream system depends on our data Would corrupted data cause a more serious operational decision? Interface boundary, dependency owner and integrity impact rationale
Service becomes mission-critical during an emergency Is availability impact still supported by the old assumptions? Recovery dependency and mission-owner explanation
Hosting or authorization boundary changes Which system and provider controls actually fall inside the boundary? Boundary diagram, inherited-control claims and approving official’s decision

The strength of a FIPS 199 assessment is not that every system receives a dramatic risk label. It is that each impact value can be traced to a defined kind of information, a credible mission consequence and an accountable decision. Calculate the objective-specific maxima correctly, carry the complete vector into the record, use the overall impact for the correct baseline and reassess when the system changes. That makes categorization a usable foundation for federal risk management rather than another ambiguous label on an asset inventory.

img