The US Department of Defense zero trust programme is the most detailed public zero trust framework in existence, and defence suppliers are increasingly expected to speak its language. It is spread across several documents: a strategy that sets goals and a deadline, a reference architecture that sets principles and pillars, a capabilities and activities catalogue that sets the work, and NSA implementation guidelines that explain how to do it. This guide explains how those pieces fit together, what the seven pillars and 45 capabilities mean in practice, how Target and Advanced levels differ, and how a programme team, inside government or in the Defense Industrial Base, can use the framework without drowning in it. It is written for security architects, identity leads and programme managers. The department now styles itself the Department of War (DoW) on its newer pages; the original documents use DoD, and this guide follows the documents.
Key takeaways
- Four documents, four jobs: the Strategy (October 2022) sets goals and the FY2027 deadline, the Reference Architecture v2.0 sets tenets and principles, the Capabilities and Activities catalogue defines the work, and the NSA Zero Trust Implementation Guidelines (ZIGs) describe how to do it.
- Seven pillars, 45 capabilities: User, Device, Applications and Workloads, Data, Network and Environment, Automation and Orchestration, and Visibility and Analytics, each broken into numbered capabilities such as 1.3 Multi-Factor Authentication and 5.4 Micro Segmentation.
- Target level is the compliance floor: components must reach Target level outcomes by the end of FY2027; Advanced level is the complete set that enables adaptive responses.
- Identity carries the most capabilities: the User pillar has nine capabilities, more than any other pillar, and the architecture requires machine-to-machine authentication, so service accounts and workloads are in scope.
- Sequence by ZIG phase, not by pillar: the Discovery phase (13 capabilities) builds the inventory everything else depends on; Phase One and Phase Two layer controls on top.
How the DoD zero trust documents fit together
Most confusion about DoD zero trust comes from treating one document as the whole framework. Each has a distinct purpose, and a programme needs all of them.
| Document | Issued by and date | What it gives you | Use it for |
|---|---|---|---|
| DoD Zero Trust Strategy v1.0 | DoD CIO, dated 21 October 2022 | Vision, four strategic goals, Target and Advanced definitions, FY2027 deadline | Executive alignment, funding narrative, governance |
| DoD Zero Trust Reference Architecture v2.0 | DISA and NSA Zero Trust Engineering Team, July 2022 | Tenets, seven principles, pillars, capability taxonomy, PDP and PEP model | Architecture decisions, investment and portfolio planning |
| Zero Trust Capabilities and Activities | DoD CIO library | 45 numbered capabilities, each with activities labelled Target or Advanced, expected outcomes | Scoping work, building a backlog, measuring progress |
| Zero Trust Capability Execution Roadmap | DoD CIO library | Timeline for activities by fiscal year | Phasing and reporting |
| Zero Trust Implementation Guidelines (ZIGs) | NSA, Primer and Discovery released 14 January 2026, then Phase One and Phase Two | Practical steps for Target level capabilities, organised by phase | Engineering teams doing the work |
| Zero Trust for Operational Technology Activities and Outcomes | DoD CIO library; NSA OT CSI 8 October 2026 | Activities adapted for OT environments | Plants, weapons systems support, facilities |
The Reference Architecture describes itself as conceptual and capability-centric: it supports capability planning and investment decisions rather than prescribing products. The ZIGs are the bridge to engineering. They are modular, so an organisation can pick the capabilities that fit its current maturity rather than working through every pillar at once.
Strategy: goals, levels and the FY2027 deadline
The Strategy sets four strategic goals. Treat them as the four workstreams of the programme, not just the technology one.
- Zero Trust Cultural Adoption: a zero trust mindset that guides design, development, integration and deployment, with all personnel aware and trained. In practice this means training for engineers and approvers, and changing how exceptions are granted.
- DoD Information Systems Secured and Defended: the controls themselves, delivered against the capabilities catalogue.
- Technology Acceleration: keeping pace with industry, including commercial and government-owned cloud enterprise services as alternative courses of action.
- Zero Trust Enablement: resourcing, governance and portfolio management. The Zero Trust Portfolio Management Office (PfMO), set up within the DoD CIO in January 2022, orchestrates the effort.
The two levels are defined precisely. Target level is the minimum set of capability outcomes and activities needed to manage risks from currently known threats. Advanced level is the complete set that enables adaptive responses to threats. Components must achieve Target level outcomes no later than the end of FY2027, with execution starting in FY23. For a supplier, the practical reading is that Target level is the bar customers will ask about; Advanced level is where automation and analytics make decisions dynamically.
The Strategy names the Defense Industrial Base as part of the DoD zero trust ecosystem and calls for communication with the DIB and allies. It does not itself place contractual obligations on suppliers; those come through contracts, DFARS clauses and CMMC. See the NIST SP 800-171 and CMMC 2.0 Guide and the DFARS 252.204-7012 Guide.
Reference Architecture: tenets and principles
The Reference Architecture v2.0 sets five tenets: assume a hostile environment, presume breach, never trust and always verify, scrutinise explicitly, and apply unified analytics, with access denied by default. It then sets seven principles that are the best test of whether a design is genuinely zero trust.
| Principle (Reference Architecture 2.4) | What it means for design | Evidence an assessor will look for |
|---|---|---|
| No network zone is trusted, explicitly or implicitly | Being on the internal network or VPN grants nothing by itself | Applications require authentication from internal ranges; no flat "trusted" VLAN rules |
| Identity-based authentication and authorisation for all access | Every request is tied to an identity, human or machine | No anonymous internal services; SSO coverage report |
| Machine-to-machine authentication and authorisation enforced | Workloads and service accounts authenticate with credentials that can be scoped and revoked | Service account inventory, mTLS or token-based workload identity, no shared static secrets |
| Near-real-time risk profiles inform user and device authorisation | Device posture and behaviour feed access decisions | Conditional access policies consuming device compliance and risk signals |
| Sensitive data encrypted in transit and at rest | Encryption follows the data, not the network segment | TLS everywhere internally, storage encryption and key management records |
| Events continuously monitored and analysed for policy compliance | Telemetry from every pillar reaches analytics | Log source coverage map, detection use cases tied to pillars |
| Policy management and distribution centralised | One policy authority, many enforcement points | Policy-as-code repository, change control on access policy |
The architecture uses the policy decision point (PDP) and policy enforcement point (PEP) model familiar from NIST SP 800-207. The PDP evaluates a confidence score for the user or endpoint, built from attributes such as identity, location, time and device posture, and there can be several decision points for authentication, authorisation, resource, application and data. The PEP enables, monitors and terminates connections, and can end an existing session when policy changes. That last point matters: if your enforcement points cannot revoke a live session, you have periodic checks, not continuous ones. See the NIST SP 800-207 Zero Trust Architecture Guide.
The seven pillars and 45 capabilities
The Capabilities and Activities catalogue numbers each capability by pillar. The counts below come from that catalogue; the examples are the capabilities most programmes start with.
| Pillar | Capabilities | Examples | Where identity teams get pulled in |
|---|---|---|---|
| 1 User | 9 (1.1 to 1.9) | 1.1 User Inventory, 1.2 Conditional User Access, 1.3 Multi-Factor Authentication, 1.4 Privileged Access Management, 1.5 Identity Federation and User Credentialing | All of it: inventory, MFA, PAM, federation, continuous authentication |
| 2 Device | 7 (2.1 to 2.7) | Device inventory, compliance and posture, endpoint detection | Device identity and certificates feeding conditional access |
| 3 Applications and Workloads | 5 (3.1 to 3.5) | Application inventory, secure software development, resource authorisation | Workload identity, service accounts, application access policy |
| 4 Data | 7 (4.1 to 4.7) | Data catalogue, labelling and tagging, data loss prevention, data access control | Attribute-based access tied to data labels |
| 5 Network and Environment | 4 (5.1 to 5.4) | 5.1 Data Flow Mapping, 5.2 Software Defined Networking, 5.3 Macro Segmentation, 5.4 Micro Segmentation | Identity-aware segmentation rules |
| 6 Automation and Orchestration | 7 (6.1 to 6.7) | 6.1 Policy Decision Point and Policy Orchestration, automated response | Central policy for access, SOAR playbooks that disable accounts |
| 7 Visibility and Analytics | 6 (7.1 to 7.6) | Log collection, analytics, user and entity behaviour analytics | Identity telemetry and anomaly detection |
Each capability is broken into activities, and each activity is labelled Target or Advanced. The pattern is consistent: Target activities establish inventory, baseline controls and integration; Advanced activities automate and make decisions adaptive. In the Network pillar, for example, Data Flow Mapping and Macro Segmentation activities are all Target level, while Software Defined Networking and Micro Segmentation each include an Advanced activity. For a pillar by pillar comparison with CISA's model, see the Zero Trust Maturity Model Guide.
Where identity fits: the User pillar and machine identities
With nine capabilities, the User pillar is the largest, and it is where most organisations find their first gaps. Three points deserve attention.
- Inventory first: 1.1 User Inventory comes before MFA and PAM because conditional access cannot be enforced on accounts nobody knows about. Include local accounts, shared accounts, break-glass accounts and accounts in cloud tenants outside the main directory.
- Phishing-resistant MFA is the realistic Target: DoD itself relies on PIV and CAC smart cards. Suppliers usually reach an equivalent assurance level with FIDO2 security keys or platform passkeys, or with certificate-based authentication. See the Smart Card and PIV/CAC Authentication Guide and the Phishing-Resistant MFA Guide.
- PAM is a User pillar capability: 1.4 Privileged Access Management means vaulting, just-in-time elevation and session control for administrators, not only a password vault. See the Privileged Access Management Guide.
- Machines are users too: the architecture principle that machine-to-machine authentication and authorisation are enforced pulls service accounts, API keys, certificates and workload identities into scope. A programme that only inventories people will fail this principle. See the Zero Trust Identity Guide.
The Change Healthcare intrusion is a clear example of what 1.2 and 1.3 are for: attackers used a stolen employee login on a remote access portal that had no multi-factor check. See the Change Healthcare breach analysis.
Implementing the framework: a phased approach
The NSA ZIGs organise Target level work into a Discovery phase (13 capabilities across all seven pillars), Phase One (30 capabilities) and Phase Two (34 capabilities). That sequencing is a sound model even for organisations outside DoD, because it forces visibility before enforcement.
- Set up governance (month 0 to 2): name an accountable executive, a programme office and pillar owners. Agree that the capabilities catalogue is the backlog and that progress is reported as activities completed with evidence.
- Discovery (months 1 to 6): build inventories of users, non-human identities, devices, applications, data stores and data flows. The Discovery ZIG is explicitly about foundational visibility into data, applications, assets, services and access activity. Do not start enforcement until the inventories are good enough to tell you what will break.
- Gap assessment: score each Target activity as not started, partial or complete, with the evidence that proves it. Record which pillar owner holds each gap.
- Phase One controls (months 4 to 18): MFA everywhere, conditional access, PAM for administrators, device compliance, macro segmentation, central logging. Deploy each in report-only or audit mode first, review the impact, then enforce in pilot rings: IT staff, then a business unit, then everyone.
- Phase Two controls (months 12 to 30): least privilege at scale, micro segmentation for high-value assets, data labelling and data-centric access control, automated response.
- Rollback and exceptions: every enforcement change needs a documented rollback (disable the policy, revert to the previous rule set) and an exception process with an owner and an expiry date. Unbounded exceptions are how implicit trust returns.
- Move towards Advanced: once Target activities are evidenced, invest in continuous authentication, behaviour analytics and automated policy orchestration.
Decision guidance: DoD framework, CISA model or NIST?
| Situation | Primary framework | Why |
|---|---|---|
| DoD component or programme office | DoD Strategy, capabilities catalogue and ZIGs | Target level by FY2027 is mandatory; reporting follows DoD activities |
| Defense Industrial Base supplier | CMMC and NIST SP 800-171 for obligations; DoD catalogue as the target architecture | Contracts drive obligations; the catalogue shows where customers are heading |
| US federal civilian agency | CISA Zero Trust Maturity Model and OMB M-22-09 | The civilian model is the expected reference; see the CISA Zero Trust Maturity Model Guide |
| Commercial organisation, any sector | NIST SP 800-207 for concepts, CISA or DoD model for measurement | The DoD catalogue is the most granular backlog available if you need one |
| Operational technology estates | DoD Zero Trust for OT activities and NSA OT guidance, alongside IEC 62443 | IT activities do not transfer directly to safety-critical systems |
Small organisations (a few hundred staff, mostly SaaS) should not try to track all 45 capabilities. Pick the Discovery and Phase One capabilities that apply, evidence them, and use the catalogue language when a defence customer asks. Large organisations benefit from adopting the catalogue numbering in their GRC tool so that evidence, risks and exceptions map to one taxonomy.
Detection and measurement
The Visibility and Analytics pillar exists because zero trust decisions are only as good as the telemetry behind them. Log and monitor at least: authentication events from every identity provider, including failures and MFA method used; conditional access decisions, including report-only results; privileged session starts and elevations; device compliance state changes; segmentation policy denies; and creation of new accounts, service principals and credentials. Watch for techniques the controls are meant to stop, such as MITRE ATT&CK T1078 Valid Accounts, T1550 Use Alternate Authentication Material and T1021 Remote Services.
- Target activity completion: percentage of applicable Target activities complete with evidence, per pillar.
- Phishing-resistant MFA coverage: share of human sign-ins using phishing-resistant methods, with privileged users reported separately.
- Inventory accuracy: accounts, devices and workloads seen in logs but missing from inventory, trending to zero.
- Non-human identity coverage: share of service accounts and workloads with an owner and scoped, rotatable credentials.
- Segmentation coverage: high-value assets behind micro segmentation policies in enforce mode.
- Exception age: number of open exceptions past their expiry date.
Framework mapping
| DoD pillar | NIST CSF 2.0 | NIST SP 800-53 Rev. 5 | CIS Controls v8.1 | ISO/IEC 27001:2022 Annex A |
|---|---|---|---|---|
| User | PR.AA-01, PR.AA-03, PR.AA-05 | AC-2, IA-2, IA-2(1), IA-5, AC-6 | 5, 6 | 5.15, 5.16, 5.17, 5.18, 8.2, 8.5 |
| Device | PR.AA-03, DE.CM-01 | IA-3, CA-7, SI-4 | 1, 4, 10 | 8.1 |
| Applications and Workloads | PR.AA-05 | AC-3, IA-5 | 16 | 8.25, 8.26 |
| Data | PR.AA-05 | AC-3 | 3 | 5.12, 8.12 |
| Network and Environment | DE.CM-01 | SC-7, AC-17 | 12, 13 | 8.20, 8.22 |
| Automation and Orchestration | PR.AA-05 | AC-3, IA-10 | 6 | 5.15 |
| Visibility and Analytics | DE.CM-01, DE.CM-03 | AU-6, SI-4, CA-7 | 8, 13 | 8.15, 8.16 |
Common failure patterns
- Buying a pillar: treating a ZTNA product or a segmentation tool as "doing zero trust". The capabilities catalogue makes clear that no single product covers a pillar, let alone the framework. See the Top 10 Zero Trust Mistakes.
- Skipping Discovery: enforcing conditional access or micro segmentation before inventories are complete, then rolling back after outages and losing executive support.
- Human-only identity scope: ignoring service accounts, API keys and workload identities, which fails the machine-to-machine principle and leaves the paths attackers use. The Snowflake breach showed how password-only accounts with no network restrictions were enough for mass data theft.
- Remote access without MFA: a single unprotected external portal undoes the User pillar, as the Change Healthcare intrusion showed.
- Counting activities without evidence: marking activities complete from configuration intent rather than test results, so the progress figure looks good and the controls do not hold.
- Permanent exceptions: legacy systems exempted indefinitely, recreating trusted zones the Reference Architecture says must not exist.
Practitioner checklist
- Read the Strategy, Reference Architecture v2.0, Capabilities and Activities catalogue and ZIG Primer together, and record which version of each you are working to.
- Appoint an owner for each of the seven pillars and a programme office that holds the backlog.
- Complete Discovery inventories for users, non-human identities, devices, applications, data and flows before enforcing.
- Score every applicable Target activity with evidence and keep the evidence with the score.
- Deploy phishing-resistant MFA and conditional access on every external entry point first.
- Bring service accounts and workload credentials into the User and Applications pillars with owners and rotation.
- Roll out each enforcement change in audit mode, then pilot rings, with a written rollback.
- Run an exception register with owners and expiry dates, and report overdue exceptions monthly.
- Map the catalogue to your existing control framework (NIST SP 800-53, 800-171 or ISO 27001) so evidence is collected once.
Standards and references
- DoD CIO: DoD Zero Trust Strategy
- DoD CIO: DoD Zero Trust Reference Architecture Version 2.0
- DoD CIO: Zero Trust Capabilities and Activities
- DoD CIO: Library (Zero Trust Capability Execution Roadmap and OT activities)
- NSA: Zero Trust Implementation Guidelines
- NSA: NSA Releases First in Series of Zero Trust Implementation Guidelines
- NIST: SP 800-53 Rev. 5
- NIST: Cybersecurity Framework 2.0
Related NHI Mgmt Group resources: Zero Trust Maturity Model Guide · NIST SP 800-207 Zero Trust Architecture Guide · CISA Zero Trust Maturity Model Guide · Zero Trust Identity Guide · CMMC 2.0 Level 2 Assessment Guide