Zero Trust Architecture (ZTA) is the security model defined in NIST Special Publication 800-207, published in August 2020 by NIST's Scott Rose and Oliver Borchert together with Stu Mitchell and DHS's Sean Connelly. The document states that zero trust moves protective measures "from static, network-based perimeters to focus on users, assets, and resources," replacing the older assumption that anything inside a network perimeter can be trusted by default. Instead, ZTA requires that "authentication and authorization (both subject and device) are discrete functions performed before a session to an enterprise resource is established" — every request is evaluated on its own merits, regardless of where it originates.
Key takeaways
- Zero Trust Architecture (ZTA) is the security model defined in NIST Special Publication 800-207, published in August 2020 by NIST's Scott Rose and Oliver Borchert together with Stu Mitchell and DHS's Sean Connelly.
- Section 2.2 adds six network assumptions that sit beside the tenets.
- The PE does not decide in a vacuum.
- NIST states that "ideally, a ZTA trust algorithm should be contextual," while noting it is not always possible with the infrastructure available.
What are the seven tenets?
Section 2.1 of SP 800-207 lists seven tenets that define the zero trust approach:
| # | Tenet |
|---|---|
| 1 | All data sources and computing services are considered resources. |
| 2 | All communication is secured regardless of network location. |
| 3 | Access to individual enterprise resources is granted on a per-session basis. |
| 4 | Access to resources is determined by dynamic policy — including the observable state of client identity, application/service, and the requesting asset. |
| 5 | The enterprise monitors and measures the integrity and security posture of all owned and associated assets. |
| 6 | All resource authentication and authorization are dynamic and strictly enforced before access is allowed. |
| 7 | The enterprise collects as much information as possible about the current state of assets, network infrastructure and communications. |
These are deliberately framed as an "ideal," not a single product configuration: SP 800-207 presents them as the goal a ZTA implementation moves toward, acknowledging that most enterprises will operate a hybrid of zero trust and legacy perimeter-based access for some time. The document also widens "resource" beyond data: its definition names "printers, compute resources, Internet of Things [IoT] actuators," which is why connected products land inside a customer's zero trust scope.
What does SP 800-207 assume about the network?
Section 2.2 adds six network assumptions that sit beside the tenets. Read from a device maker's side, they describe the environment your product will be installed into:
- The enterprise private network is not an implicit trust zone, which "entails actions such as authenticating all connections and encrypting all traffic."
- Devices on the network may not be owned or configurable by the enterprise.
- No resource is inherently trusted: every asset's posture is evaluated via a PEP, and "subject credentials alone are insufficient for device authentication."
- Not all enterprise resources are on enterprise-owned infrastructure.
- Remote subjects and assets cannot fully trust their local network connection.
- Assets moving between enterprise and non-enterprise infrastructure keep a consistent policy and posture.
Section 3.4.1 then lists ten network requirements. Two matter most for products: the enterprise must tell owned assets apart using "enterprise-issued credentials," not unauthenticated data such as "network MAC addresses that can be spoofed," and enterprise resources "should not be reachable without accessing a PEP."
Which logical components make the access decision?
Section 3 of SP 800-207 splits the policy decision point into two components and adds an enforcement point:
| Component | Role |
|---|---|
| Policy Engine (PE) | "Responsible for the ultimate decision to grant access to a resource for a given subject." Uses enterprise policy plus input from external sources (identity management, threat intelligence, SIEM data) to compute a trust score or confidence level. |
| Policy Administrator (PA) | "Responsible for establishing and/or shutting down the communication path between a subject and a resource" — it executes the PE's decision by generating any session-specific credential or token the client uses to access the resource. |
| Policy Enforcement Point (PEP) | "Responsible for enabling, monitoring, and eventually terminating connections between a subject and an enterprise resource." This is the single logical component that sits in the data path itself; it may be split into client-side and resource-side parts. |
The PE does not decide in a vacuum. The same section lists the data sources that feed it:
| Data source | What it tells the PE |
|---|---|
| Continuous diagnostics and mitigation (CDM) | Asset state, approved software, "whether the asset has any known vulnerabilities" |
| Threat intelligence feed(s) | New attacks, vulnerabilities and malware |
| Enterprise PKI and ID management | Certificates, accounts, roles, assigned assets |
| Activity logs and SIEM | Near-real-time posture; input to refine policy |
| Compliance system and data access policies | Regulatory rules; baseline privileges by role |
For a device maker, the CDM row is the important one: a customer's PE compares what your device reports (firmware version, components) against what it knows about vulnerabilities in that build.
How does the trust algorithm decide?
Section 3.3 calls the trust algorithm "the process used by the policy engine to ultimately grant or deny access to a resource." Its inputs are the access request, the subject database, the asset database and observable status (OS version, software present and its integrity, location, patch level), resource requirements, and threat intelligence. Section 3.3.1 then describes two independent design choices:
| Choice | Option A | Option B |
|---|---|---|
| How factors are evaluated | Criteria-based: a set of qualified attributes must all be met before access is granted | Score-based: a confidence level computed from every data source and enterprise-set weights, compared with a per-resource threshold; below it, access is denied or privileges reduced |
| How requests relate to each other | Singular: each request evaluated alone, faster, but an attack that stays within a subject's allowed role can go unnoticed | Contextual: recent history of the subject or agent is considered, which can catch subverted credentials used in an atypical pattern |
NIST states that "ideally, a ZTA trust algorithm should be contextual," while noting it is not always possible with the infrastructure available. A device that cannot report its own state forces the customer toward criteria checked at the network edge rather than a richer score.
Which deployment models does SP 800-207 describe?
SP 800-207 does not mandate one architecture. Section 3.1 names three approaches that can drive policy (enhanced identity governance, micro-segmentation, and network infrastructure / software defined perimeters), and notes that "a full ZT solution will include elements of all three." Section 3.2 then gives four deployment variations an enterprise can combine:
| Model | How it works | Trade-off SP 800-207 notes |
|---|---|---|
| Device agent/gateway | An agent on the client device coordinates with a gateway PEP in front of each resource | Suits enterprises with "a robust device management program"; needs an agent on every asset |
| Enclave-based | A gateway PEP sits in front of a group of resources, useful for legacy systems that cannot support their own agent | The gateway "may not be able to protect each resource individually" |
| Resource portal | A single portal acts as the PEP; no client agent | Assets are only visible when they connect; the portal is a DoS target |
| Application sandboxing | Access happens through an isolated application on the client device | Every sandboxed application must be maintained |
Most embedded devices, controllers and cameras cannot host a vendor's ZT agent. In practice they end up in the enclave model: a gateway in front of a device segment, with the device's own identity and posture as inputs where the device can provide them. The cloud and multi-location variant of this design is covered in SP 800-207A.
What does NCCoE SP 1800-35 add?
NIST SP 1800-35, "Implementing a Zero Trust Architecture," published June 2025, is the NCCoE practice guide that turns SP 800-207 into lab builds. Per its CSRC abstract, the NCCoE "worked with 24 collaborators" to "build 19 ZTA example implementations." The high-level document groups the builds into enhanced identity governance "crawl" (on-premises only) and "run" (adding cloud resources and device discovery) phases, followed by software-defined perimeter, microsegmentation and secure access service edge (SASE) builds. It also maps the example capabilities to CSF 1.1, CSF 2.0 and SP 800-53 Rev. 5, and states plainly that ZTA "is a set of concepts and principles, not a set of technical specifications that can be complied with."
Its scope note is the line device makers should read twice: "ZTA for industrial control systems, operational technology (OT) environments, and Internet of Things (IoT) devices are explicitly out of scope." The seven "journey takeaways" in section 8 still apply to a plant or a fleet of devices:
- Discover and inventory the existing environment.
- Formulate access policy to support the mission and business use cases.
- Identify existing security capabilities and technology.
- Eliminate gaps in zero trust policy and processes by applying a risk-based approach based on the value of data.
- Implement ZTA components (people, process, and technology) and incrementally use deployed security solutions.
- Verify the implementation to support zero trust outcomes.
- Continuously improve and evolve due to changes in threat landscape, mission, technology, and regulations.
Which threats does a ZTA itself introduce?
Section 5 of SP 800-207 is explicit that "no enterprise can eliminate cybersecurity risk" and lists threats specific to running a ZTA:
| § | Threat |
|---|---|
| 5.1 | Subversion of the ZTA decision process (PE/PA configuration changes must be logged and audited) |
| 5.2 | Denial-of-service or network disruption (NIST cites the Mirai IoT botnet) |
| 5.3 | Stolen credentials / insider threat |
| 5.4 | Visibility on the network |
| 5.5 | Storage of system and network information |
| 5.6 | Reliance on proprietary data formats or solutions |
| 5.7 | Use of non-person entities (NPE) in ZTA administration |
For OT sites, 5.2 is the one to plan for first: a device that loses its PEP path must fail in a way the process can tolerate.
How does CISA's Zero Trust Maturity Model relate to SP 800-207?
NIST's SP 800-207 defines the architecture; it does not tell an organization how mature its own implementation is or prioritize what to build first. CISA's Zero Trust Maturity Model (version 2.0, published April 2023) fills that gap for US federal civilian agencies, and CISA's own resource page states it aligns to OMB memorandum M-22-09. It organizes the same underlying concept into five pillars plus three cross-cutting capabilities, each scored across four maturity stages:
| Pillar / capability | |
|---|---|
| Identity, Devices, Networks, Applications and Workloads, Data | the five pillars |
| Visibility and Analytics, Automation and Orchestration, Governance | the three cross-cutting capabilities |
| Stage | CISA's description |
|---|---|
| Traditional | Manually configured lifecycles and attributes; static security policies; siloed pillars; manual response. |
| Initial | Starting automation of attribute assignment, configuration and enforcement; initial cross-pillar integration. |
| Advanced | Automated lifecycle and configuration controls with cross-pillar coordination; centralized visibility and identity control. |
| Optimal | Fully automated, just-in-time lifecycles; dynamic least-privilege access; cross-pillar interoperability with continuous monitoring. |
CISA's model is specific to the US federal civilian executive branch; it is not a legal requirement for device makers, OT asset owners or software vendors elsewhere, though both documents are widely cited as the reference vocabulary for "zero trust" regardless of sector.
What does a zero trust customer ask of a connected device?
ZTA was written with enterprise IT networks in mind, not embedded devices, but its tenets translate directly to device fleets and industrial networks: per-session (not per-perimeter) authentication for every device-to-service connection, continuous posture checks rather than a one-time boot-time trust decision, and dynamic policy that can react to a device's observed state — exactly the kind of evidence a firmware security assessment or passive OT network snapshot produces.
The table below is one reading of SP 800-207 from the product side. NIST does not set device requirements here; these are the questions a customer's architects tend to derive from the text.
| SP 800-207 anchor | What the customer's PE or PEP needs | What the device maker can ship |
|---|---|---|
| Tenet 2, network assumption 1 | Authenticated, encrypted traffic on every interface | TLS or equivalent on management and data channels; no clear-text admin protocols on by default |
| Network requirement 2 | Identity from enterprise-issued credentials, not MAC addresses | Per-device certificate or credential slot the operator can provision; see SP 800-63 for assurance levels on the human side |
| Tenet 5, CDM data source | Firmware version, components and known vulnerabilities | A readable firmware version and a firmware SBOM per release |
| Network requirement 4 | Resources reachable only through a PEP | Documented ports and flows so a gateway can allow only those; a MUD file is one machine-readable way |
| Threat 5.2 | Predictable behaviour when the PEP path drops | Documented fail-safe state and reconnect behaviour |
Worked example: how does a policy engine treat one camera fleet?
Synthetic example. The site, product, versions and rules are invented to show how SP 800-207's pieces combine; they are not NIST values and do not describe any real product.
A logistics site places 40 IP cameras from one vendor in an enclave behind a gateway PEP (section 3.2.2). The video management server asks for a stream session to each camera. The PE uses a criteria-based trust algorithm with these criteria for the camera enclave:
| Input | Criterion | Camera A (fw 4.2.1) | Camera B (fw 4.1.0) |
|---|---|---|---|
| Asset database | Device certificate issued by site PKI | Present | Present |
| Asset status | Firmware on approved list | Yes | Yes |
| Threat intelligence | No KEV-listed CVE matched to this firmware's components | None matched | One matched in the bundled web server |
| Resource requirement | Session only from the video server, port 554 | Met | Met |
| Decision | All criteria must hold | Stream allowed | Stream allowed only via read-only path; admin web UI blocked; ticket to vendor |
The decision on camera B rests on two vendor artifacts: a component list precise enough to match against the CISA KEV catalog and EPSS scores, and a fixed firmware the site can roll out. Without either, the site's only lever is to isolate the whole enclave.
What are the common pitfalls?
- Treating ZTA as a product. "Our device is zero trust certified" has nothing to point to; SP 1800-35 says ZTA cannot be complied with as a specification.
- Using MAC addresses as device identity. Section 3.4.1 rules this out.
- Assuming an agent will run on the device. Most embedded targets fit the enclave model.
- One-time trust at boot. Tenet 5 expects continuing posture measurement.
- Ignoring the PE/PA as a single point of failure. Threat 5.2 weighs more in OT, where an access outage can stop a process.
- Expecting SP 1800-35 to cover devices. It puts ICS, OT and IoT out of scope.
SoftWSP's device and firmware assessments produce the kind of per-asset posture evidence (firmware version, components from an extracted SBOM, known CVEs matched against NVD, CISA KEV, FIRST EPSS and CISA ICS advisories) a Policy Engine needs as an input, in a cited report an engineer reviews, rather than a one-time compliance checkbox: see device firmware security.
This guide summarises public NIST and CISA publications; it does not constitute legal or compliance advice.
Put this guide to work
Frequently asked questions
What is Zero Trust Architecture according to NIST?
NIST Special Publication 800-207, published in August 2020, defines zero trust as a set of security paradigms that move protection "from static, network-based perimeters to focus on users, assets, and resources." A Zero Trust Architecture (ZTA) is an enterprise's cybersecurity plan built on those paradigms: it requires authentication and authorization of both the subject and the device as discrete steps performed before any session to an enterprise resource is established, and it focuses on protecting individual resources rather than network segments.
What are NIST's seven tenets of zero trust?
Section 2.1 of SP 800-207 lists seven: (1) all data sources and computing services are considered resources; (2) all communication is secured regardless of network location; (3) access to individual enterprise resources is granted on a per-session basis; (4) access is determined by dynamic policy, including the observable state of client identity, application/service, and the requesting asset; (5) the enterprise monitors and measures the integrity and security posture of all owned and associated assets; (6) all resource authentication and authorization are dynamic and strictly enforced before access is allowed; and (7) the enterprise collects as much information as possible about the current state of assets, network infrastructure and communications.
What are the policy engine, policy administrator and policy enforcement point?
The three core logical components SP 800-207 uses to describe how a ZTA makes and carries out access decisions. The Policy Engine (PE) is "responsible for the ultimate decision to grant access to a resource for a given subject." The Policy Administrator (PA) is "responsible for establishing and/or shutting down the communication path between a subject and a resource" based on the PE's decision. The Policy Enforcement Point (PEP) is "responsible for enabling, monitoring, and eventually terminating connections between a subject and an enterprise resource," and is the component that actually sits in the data path.
How does CISA's Zero Trust Maturity Model relate to SP 800-207?
They are separate documents from separate agencies that describe the same underlying concept from different angles. NIST's SP 800-207 defines the architecture's logical components and tenets; CISA's Zero Trust Maturity Model (version 2.0, published April 2023, which CISA's own page states aligns to OMB memorandum M-22-09) gives US federal agencies a five-pillar (Identity, Devices, Networks, Applications and Workloads, Data), four-stage (Traditional, Initial, Advanced, Optimal) maturity scale for adopting it. Neither document binds organizations outside the US federal government; both are widely used as a voluntary reference elsewhere.
Sources
- NIST: SP 800-207, Zero Trust Architecture (final, August 2020) — resource page
- NIST: SP 800-207, Zero Trust Architecture (full text PDF)
- CISA: Zero Trust Maturity Model — resource page
- CISA: Zero Trust Maturity Model, Version 2.0 (full text PDF, April 2023)
- NIST: SP 1800-35, Implementing a Zero Trust Architecture: High-Level Document (June 2025) — publication page
- NIST SP 1800-35, Implementing a Zero Trust Architecture (high-level document PDF)
About this guide
SoftWSP Firmware Security Desk
Covers the Firmware vulnerabilities & fixes cluster: 129 guides.

In this cluster
Questions about this page?
