Skip to content
StratQuad

NCSC Cyber Assessment Framework (CAF)

5 min read

The NCSC's outcome based framework for assessing how well an organisation manages cyber risk to the functions it must keep running.

At a glance

Owner
National Cyber Security Centre (NCSC)
Version
CAF 4.0
Published
6 Aug 2025
Applies to
Operators of essential services, UK government departments, NHS bodies, local government and other organisations whose regulator or sponsor adopts it
Mandatory
Mandatory where a regulator or sponsor requires it (NIS Regulations, GovAssure, DSPT); voluntary otherwise
Assessed by
Self assessment, reviewed by the regulator or sponsor; independent assessment under GovAssure and DSPT
Cost model
Free to use

Key points

  • Four objectives, 14 principles and 41 contributing outcomes.
  • Each contributing outcome is rated Achieved, Partially achieved or Not achieved, using Indicators of Good Practice as aids to judgement, not a checklist.
  • CAF 4.0, published 6 August 2025, adds outcomes on understanding the threat (A2.b) and secure software development and support (A4.b), and reworks detection under C1.
  • Regulators and sponsors set which outcomes must be achieved through CAF profiles. The framework itself sets no pass mark.
  • It is the assessment basis for the NIS Regulations, GovAssure and the CAF aligned Data Security and Protection Toolkit.

What it is

The CAF assesses whether an organisation is managing cyber risk to its essential functions: the services or operations whose disruption would cause serious harm. It is written as outcomes rather than controls. It states what good looks like and leaves the organisation to choose how to achieve it.

It was created for the Network and Information Systems (NIS) Regulations 2018 and is now used well beyond them: across central government through GovAssure, by NHS England through the CAF aligned Data Security and Protection Toolkit, and by local government.

Structure

The framework has four objectives. Each objective contains principles, and each principle contains contributing outcomes. Contributing outcomes are the unit of assessment.

  • AManaging security risk
    • A1Governance 3
      • A1.aBoard Direction
      • A1.bRoles and Responsibilities
      • A1.cDecision making
    • A2Risk Management 3
      • A2.aRisk Management Process
      • A2.bUnderstanding Threat
      • A2.cAssurance
    • A3Asset Management 1
      • A3.aAsset Management
    • A4Supply Chain 2
      • A4.aSupply Chain
      • A4.bSecure Software Development and Support
  • BProtecting against cyber attack
    • B1Service Protection Policies, Processes and Procedures 2
      • B1.aPolicy, Process and Procedure Development
      • B1.bPolicy, Process and Procedure Implementation
    • B2Identity and Access Control 4
      • B2.aIdentity Verification, Authentication and Authorisation
      • B2.bDevice Management
      • B2.cPrivileged User Management
      • B2.dIdentity and Access Management
    • B3Data Security 5
      • B3.aUnderstanding Data
      • B3.bData in Transit
      • B3.cStored Data
      • B3.dMobile Data
      • B3.eMedia and Equipment Sanitisation
    • B4System Security 4
      • B4.aSecure by Design
      • B4.bSecure Configuration
      • B4.cSecure Management
      • B4.dVulnerability Management
    • B5Resilient Networks and Systems 3
      • B5.aResilience Preparation
      • B5.bDesign for Resilience
      • B5.cBackups
    • B6Staff Awareness and Training 2
      • B6.aCyber Security Culture
      • B6.bCyber Security Training
  • CDetecting cyber security events
    • C1Security Monitoring 6
      • C1.aSources and Tools for Logging and Monitoring
      • C1.bSecuring Logs
      • C1.cGenerating Alerts
      • C1.dTriage of Security Alerts
      • C1.ePersonnel Skills for Monitoring and Detection
      • C1.fUnderstanding User and System Behaviour, and Threat Intelligence
    • C2Threat Hunting 1
      • C2.aThreat Hunting
  • DMinimising the impact of cyber security incidents
    • D1Response and Recovery Planning 3
      • D1.aResponse Plan
      • D1.bResponse and Recovery Capability
      • D1.cTesting and Exercising
    • D2Lessons Learned 2
      • D2.aPost Incident Analysis
      • D2.bUsing Incidents to Drive Improvements

How outcomes are rated

Each contributing outcome has a table of Indicators of Good Practice (IGPs) in up to three columns:

RatingWhat the IGPs describeHow the CAF says to use them
AchievedTypical characteristics of full achievementNormally all are true, unless compensating measures meet the outcome another way
Partially achievedPartial achievement that still delivers a specific, worthwhile security benefitNot available for every outcome
Not achievedTypical characteristics of failureAny one is normally enough to justify this rating

The CAF is explicit that IGPs support expert judgement and are not an exhaustive or rigid checklist. An assessor can rate an outcome Achieved where the organisation meets its intent by a route the IGPs do not describe.

What changed in CAF 4.0

Changes from CAF 3.2 to CAF 4.0

  • A2.b Understanding Threat (new). Requires an understanding of threat actors' capabilities and intent against the essential function. Threat analysis indicators move here from A2.a, and Assurance is renumbered from A2.b to A2.c.
  • A4.b Secure Software Development and Support (new). Covers secure, supported software, including provenance, testing, patching and attestation of authenticity, whether developed in house or by suppliers.
  • A4.a Supply Chain. Expanded indicators on supplier security and contractual obligations.
  • Detection reworked. C1.a becomes Sources and Tools for Logging and Monitoring, C1.d becomes Triage of Security Alerts, C1.e becomes Personnel Skills for Monitoring and Detection, and threat intelligence indicators move into C1.f.
  • Automated decision making. A new indicator under B4.a addresses restrictions on automated decision making technologies that could adversely affect essential functions. NCSC describes this as broader coverage of AI related risk.
  • Language. "Attacker" becomes "threat actor" across many outcomes, and descriptions refer to security and resilience.

Profiles and who sets the bar

The CAF does not say how many outcomes must be Achieved. That is set by whoever uses it:

  • NIS competent authorities set expectations for operators of essential services in their sector.
  • GovAssure defines Baseline and Enhanced CAF profiles for government departments, chosen by the criticality of each system.
  • NHS England maps the CAF to the Data Security and Protection Toolkit and specifies which outcomes each organisation type must meet.

Practitioner note

Before any assessment, get the profile in writing. A Partially achieved rating on an outcome the profile expects to be Achieved is a gap to remediate; the same rating on an outcome outside the profile may be entirely acceptable. Assessments that skip this step tend to produce long lists of "gaps" the organisation was never required to close.

How it is assessed

  1. Define the scope. Identify the essential functions and the networks and information systems they depend on. CAF ratings attach to that scope, not to the whole organisation.
  2. Self assess. For each contributing outcome, rate against the IGPs and record evidence.
  3. Review or independent assessment. Depending on the regime, the regulator, an independent assessor or both review the self assessment.
  4. Improvement plan. Gaps against the profile go into a plan with owners and dates, which the regulator or sponsor tracks.

Evidence expected

Assessors look for operating evidence, not policy alone: board minutes showing cyber risk decisions (A1), a current threat assessment for the essential function (A2.b), an asset inventory that includes dependencies and data flows (A3.a), supplier security requirements in contracts (A4.a), privileged access reviews (B2.c), patching metrics against defined timescales (B4.d), tested backup restores (B5.c), alert triage records (C1.d), and exercise reports with tracked actions (D1.c).

Common gaps

Common nonconformity

  • Essential functions not defined, so scope is "everything" and evidence never lines up with ratings.
  • A2.b rated Achieved on the strength of a subscription to a threat feed, with no analysis of what it means for the essential function.
  • Logging in place but no defined triage process or skilled staff to act on alerts, which now fails C1.d and C1.e explicitly.
  • Backups that exist but have never been restored under test conditions (B5.c).
  • Incident response plans that are not exercised, or exercises whose lessons are never tracked to closure (D1.c, D2.b).

Relationship to other frameworks

The CAF overlaps heavily with ISO/IEC 27001 Annex A and the NIST Cybersecurity Framework, but is not interchangeable with either. ISO/IEC 27001 certifies a management system; the CAF assesses outcomes for specific essential functions; NIST CSF is a voluntary outcome taxonomy without an assessment scheme. Cyber Essentials covers a small subset of principles B2 and B4. For control level mappings, use OpenCRE.

Sources

Suggest a correction