Give Your CSF Profile a Threat Axis
CSF 2.0's six functions say what you do about cyber risk. The Core deliberately leaves open against what. Cross the functions with ten cause clusters and you get 60 named control objectives — and every gap in your coverage becomes a named, empty cell.
Bernhard Kreinz · updated for TLCTC v2.6, September 26, 2026 (first published June 2025)
Functions are verbs. Your threat register needs nouns.
CSF 2.0 organises outcomes by function, not by threat type — by design, so it fits any organisation. But its Risk Assessment category (ID.RA) still asks you to identify threats, and the supporting guidance defines a threat very broadly. NIST SP 800-30: “Any circumstance or event with the potential to adversely impact organizational operations and assets, individuals, other organizations, or the Nation through an information system via unauthorized access, destruction, disclosure, or modification of information, and/or denial of service.” So every organisation fills the threat axis with its own vocabulary — and control mappings become neither comparable across organisations nor testable within one.
Six functions, 22 categories, a shared outcome language. Deliberately threat-agnostic.
Ten non-overlapping clusters, each defined by the generic vulnerability it exploits — a cause, not an outcome. “Ransomware” and “data breach” are consequences; the clusters name what made them possible.
10 causes × 6 functions = 60 control objectives
For each cluster, define a control objective under each function. What changes by column is the type of objective; what changes by row is the cause. Each cell is realised at two scopes — a local control (asset or system specific) and an umbrella control (enterprise-wide, shared). Click any cell to see which of the 106 CSF 2.0 subcategories land in it.
Objectives follow the template of the application paper §8.1; generic vulnerabilities are quoted from the v2.6 dictionary. CSF subcategory text is verbatim from NIST CSWP 29, Appendix A; each sits in its own function column, and which rows it lands in is TLCTC starter guidance (mapping with rationales). Two subcategories have no attacker cause and sit in no cell: GV.RM-07 (positive risks) and PR.IR-02 (environmental threats). The ISO/IEC 27001:2022 Annex A controls are starter guidance from the TLCTC Control Matrix tool. Both sets are AI-assisted, not an official or certified mapping. Replace them with your own.
#2 Exploiting Server, across all six functions
The same exercise, done for each of the ten clusters, is your control plan. Here is the application paper's worked example.
| Function | Control objective | Local controls | Umbrella controls |
|---|---|---|---|
| GOVERN | Govern strategy and ownership for server-exploitation risk | AppSec / server-owner accountability, secure-SDLC policy, patch-SLA standard | risk-register entry for #2 with appetite and tolerance, vulnerability-management program (ISMS clause) |
| IDENTIFY | Identify failures in the code of the server software | fuzz testing, network-based vulnerability scanning | threat intelligence, CVE subscriptions |
| PROTECT | Protect the server from being exploited | patch management, secure coding | web application firewall (WAF) |
| DETECT | Detect an exploited server | local event logging | security information and event management (SIEM) |
| RESPOND | Respond to an exploited server | emergency patching, CSIRT engagement | exploited-server response plan (e.g., enterprise WAF-rule rollout) |
| RECOVER | Recover from a server-exploit event | maintained repositories, restore procedures | IT service continuity management (IT-SCM) |
Notice what the row does for a post-incident review: when a server was exploited, you know exactly which six objectives to check — and the same row applies to every #2 incident, in every business unit, every year.
Plugs into the categories you already use
TLCTC does not add a seventh function or replace a category. It gives the existing ones a threat axis.
Record threats as clusters and typical paths — #9 → #4 → #1 — instead of outcome labels. The application paper adopts TLCTC as the ID.RA threat taxonomy.
State risk appetite per cluster row, and derive detection and containment targets from it per velocity class.
Name one owner per row — the GOV-local cell — accountable for all six objectives of that cause.
#10 sits at the Trust Acceptance Event: the moment a supplier's artifact becomes authoritative in your domain. A vulnerable dependency alone is not a supply-chain attack.
Mostly the #4 row — credential use at the point of authentication — plus authorisation scope, which #1 abuses.
The #9 row — and the reason to review controls per path: a #9 step can walk around excellent #4 controls.
Patching, hardening and execution control serve the #2, #3 and #7 rows; capacity and availability tolerance serve #6.
Every row — but judged against attack speed: detect and contain a step before the next one starts (see below).
A Current and a Target Organizational Profile become two states of the same 60 cells. The gap analysis is a list of named cells, not a narrative.
GOVERN is the management-system layer
The other five functions act on the threat. GOVERN is where each cluster plugs into your management system: ISO/IEC 27001 clauses 4–10 (context, leadership, planning, support, operation, performance evaluation, improvement) and the NIST RMF (SP 800-37 / SP 800-39: risk frame, tolerance, control ownership, the link into the enterprise risk register).
Only the GOV-umbrella cells are cross-cutting — the integration layer to enterprise risk management. GOV-local (owner, cluster policy, exceptions) and every other cell stay cluster-specific.
| Cluster | GOVERN anchors |
|---|---|
| #1 Abuse of Functions | feature/abuse-case review policy, business-logic risk acceptance |
| #2 Exploiting Server | secure-SDLC policy, patch SLA, vulnerability-management program |
| #3 Exploiting Client | endpoint-hardening standard, browser/e-mail client governance |
| #4 Identity Theft | IAM policy, MFA mandate, IAM governance committee, IGA program |
| #5 Man in the Middle | cryptographic standard, TLS policy, PKI / key-management governance |
| #6 Flooding Attack | service-availability tolerance, DDoS-protection program, capacity governance |
| #7 Malware | anti-malware/EDR policy, FEC-handling rules, enterprise malware-defense program |
| #8 Physical Attack | site/asset access policy, physical-security program |
| #9 Social Engineering | acceptable-use and awareness policy, security-awareness program governance |
| #10 Supply Chain Attack | vendor-onboarding policy, SBOM mandate, third-party risk management (TPRM) program |
A four-hour MTTD is excellent — or useless
It depends on how fast the attacker moves from one step to the next (Δt). The Detection Coverage Score puts the two times side by side, read at the 90th percentile of your own times:
Below 1 you act before the attacker completes the step; above 1 the step completes first. Only DCS_c < 1 stops it — detecting in time but containing too late means the step was seen and completed anyway.
That gives each DETECT and RESPOND cell a key control indicator, with a target derived from risk appetite. Key risk indicators sit on the rows (events and near-misses per cluster); key control indicators on the cells.
Test your times in the Δt Race| Velocity class | Typical Δt | DCS target | Required TTD_P90 / TTC_P90 | Investment focus |
|---|---|---|---|---|
| Strategic (VC-1) | 7 days | ≤ 0.5 | < 3.5 days | Threat-hunting cycles |
| Tactical (VC-2) | 4 hours | ≤ 0.8 | < 3.2 hours | SOC SLA, alert tuning |
| Operational (VC-3) | 10 min | ≤ 0.8 | < 8 min | Automation, playbooks |
| Real-Time (VC-4) | 30 sec | N/A | N/A | Prevention only |
Example targets from the application paper §10.3. Below about a minute, no detection is fast enough: the answer is architecture that denies the step.
Four steps from CSF Profile to cause-based control plan
Rewrite your threat register as clusters and typical paths.
Start from the NIST CSF 2.0 or ISO 27001 starter, set a target maturity, name a row owner.
Per velocity class, from risk appetite — then race real attack paths against your times.
Walk each step through its row: which objective should have prevented, caught or contained it?
FAQ
Does TLCTC replace the CSF?
No. The functions and categories stay exactly as they are. TLCTC supplies the threat axis the Core leaves open: clusters are nouns (causes), functions are verbs (responses). They answer different questions and never compete to classify the same thing.
Is this an official NIST mapping?
No. TLCTC is an independent, openly licensed (CC BY 4.0) framework. CSF 2.0 leaves the choice of threat taxonomy to the organisation; this page shows how TLCTC fills that choice.
Where do the 106 CSF subcategories land?
Each stays in its own function column; only the row is new. 32 address the generic vulnerability of specific clusters and land in those rows as local outcomes — PR.AA-03 authentication in #4, PR.PS-05 execution prevention in #7, all ten GV.SC outcomes in #10. 67 are cause-neutral and serve every row as umbrella outcomes, among them every RESPOND subcategory. Five act one layer later, at the Data Risk Event — data-at-rest and data-in-use protection, backups, backup verification and data inventories (PR.DS-01, PR.DS-10, PR.DS-11, RC.RP-03, ID.AM-07) — and sit in the separate DRE Control Matrix. That split is the point: the Core names cause-specific outcomes mostly in IDENTIFY and PROTECT, and leaves DETECT, RESPOND and RECOVER cause-agnostic. The rows add the cause there. The Control Matrix loads the 101 cluster-layer outcomes as a starter, the DRE Matrix the five data-layer ones.
How does ISO/IEC 27001 fit in?
As a complement. Annex A (and ISO/IEC 27002) supplies controls for the cells; ISO/IEC 27005 supplies the risk method; clauses 4–10 are where the GOVERN cells live. The Control Matrix tool ships an ISO 27001:2022 starter for all 60 cells.
Why not just use MITRE ATT&CK as the threat axis?
ATT&CK catalogues hundreds of techniques — the right level for detection engineering, too fine for a control plan or a board. Every technique maps to the clusters it uses (see the ATT&CK Explorer), so the two work together: techniques inside the cells, clusters on the rows.
What about data breaches and ransomware?
They are outcomes, not causes, so they are not rows. A ransomware incident is a path — typically ending #1 → #7 — with data made unusable as the recorded consequence. Its controls sit in the rows of the steps that made it possible.
Where does an insider fit?
Who the attacker is never decides the row. An insider acting outside their granted rights is classified like anyone else; an authorised person misusing rights they were given is Abuse of Rights — operational risk with a named owner, not a cyber cluster.
Go deeper
Sources: NIST, The NIST Cybersecurity Framework (CSF) 2.0, CSWP 29, February 2024; NIST SP 800-30 Rev. 1, Guide for Conducting Risk Assessments; NIST SP 800-37 Rev. 2 and SP 800-39; ISO/IEC 27001:2022. TLCTC application paper (v2.6) §8–10; TLCTC framework dictionary v2.6.