- Zero Trust is an architecture and a set of principles, not a product. It removes implicit trust based on network location alone and decides access from identity, device state, context and risk, with least privilege and an assumption that a breach can happen.
- NIST SP 800-207 (2020) sets out the tenets and the logical components that make access decisions. CISA's Zero Trust Maturity Model v2.0 (2023) organises the work into five pillars and three cross-cutting capabilities.
- Many organisations start with identity because it is a practical first step that uses tools they already licence. Your own risk assessment should set the order.
- Microsoft 365 Business Premium includes Conditional Access, Intune and Defender for Business. Privileged Identity Management and risk-based sign-in policies need Entra ID P2 or an equivalent add-on.
- Zero Trust controls support NIS, NIS2, ISO 27001 and DORA requirements, but they do not show conformity on their own. You still need the policies, risk assessments and evidence each framework asks for.
What Zero Trust architecture is
Zero Trust architecture is a security model in which no user, device or workload is trusted just because of where it connects from. Each request for a resource is checked against policy using the identity behind it, the state of the device, and signals such as location, time and behaviour. Access is limited to what the task needs. John Kindervag described the model at Forrester Research in 2010, and NIST Special Publication 800-207 formalised it in 2020.
Why the perimeter stopped working
Traditional network security rested on one assumption: traffic inside the corporate network is trusted, traffic outside is not. A firewall held the line and remote workers tunnelled in through a VPN. That worked when most users sat in the same building, applications were run on infrastructure you owned, and data lived on servers you could physically touch.
Today users work from home and client offices, applications run in Microsoft 365, Salesforce, AWS and dozens of SaaS platforms, and partners connect from networks you have never seen. An attacker who compromises one laptop, one phished account or one VPN credential lands inside the trusted zone, where internal traffic often goes uninspected and lateral movement can be quick.
Zero Trust removes implicit trust based on network location alone. Location can still be one signal in a policy (a sign-in from a country where you have no staff is worth blocking), but it no longer grants access by itself. Depending on the architecture, decisions are made per session or per request and re-evaluated as policy, context and risk signals change.
How access decisions are made
Section 3 of NIST SP 800-207 describes three logical components. The policy engine (PE) decides whether to grant access, using policy and inputs such as identity, device posture and threat intelligence. The policy administrator (PA) acts on that decision by setting up or closing the session. The policy enforcement point (PEP) sits in the path between the user and the resource and allows, blocks or ends the connection. Together the PE and PA form the policy decision point.
In a Microsoft 365 environment, Entra ID Conditional Access plays the decision role for cloud sign-ins, and the application, proxy or ZTNA connector enforces it. Products differ in how they split these roles, so map your own tools to the components rather than assuming one product covers all three.
The five pillars
CISA's Zero Trust Maturity Model v2.0 groups the work into five pillars: Identity, Devices, Networks, Applications and Workloads, and Data. Three cross-cutting capabilities run across all of them: Visibility and Analytics, Automation and Orchestration, and Governance. The model describes maturity stages for each pillar, from traditional to optimal.
Starting with identity
Identity is a common starting point because it covers every cloud application at once, uses tools many organisations already licence and needs no network changes. It is a practical choice, not a rule: if your main exposure is unpatched internet-facing systems or a flat OT network, start there.
Conditional Access policies
Conditional Access assesses each sign-in against signals such as user, device state, location, target application and, where licensed, risk scores from Entra ID Protection. The policy then allows, blocks or demands step-up verification. A baseline set requires MFA for all users, blocks legacy authentication, blocks sign-ins from countries where you have no presence and requires a compliant device for sensitive applications. Policies that react to user risk or sign-in risk need Entra ID P2.
Privileged Identity Management
Admin accounts are the highest-value target in any environment. PIM removes standing privilege: an administrator requests, justifies and activates a role for a limited time. Between activations the account holds no admin rights, so a stolen password alone does not give an attacker admin access. Someone who can also pass MFA and any approval step could still activate the role, so protect activation with phishing-resistant MFA and require approval for the most sensitive roles. PIM needs Entra ID P2 or Entra ID Governance.
What Microsoft 365 Business Premium includes
| Capability | In Business Premium? | Notes |
|---|---|---|
| Conditional Access (Entra ID P1) | Included | Covers MFA, device compliance and location policies |
| Intune device management | Included | Compliance policies feed Conditional Access |
| Defender for Business (EDR) | Included | Endpoint detection and response for up to 300 users |
| Privileged Identity Management | Not included | Needs Entra ID P2 or Entra ID Governance |
| Risk-based policies (user and sign-in risk) | Not included | Needs Entra ID P2 |
| Full Defender for Cloud Apps (CASB) | Not included | Separate licence or add-on |
Microsoft changes its plans from time to time, so check the Business Premium service description before you buy.
Device trust and compliance
A valid MFA sign-in from a laptop running unpatched software or carrying an active infection is still a real access risk. Compliance policies in Intune or Jamf define the minimum bar: operating system version, encryption, EDR agent, screen lock and jailbreak detection on mobiles. The endpoint security guide covers each of these controls. Conditional Access checks compliance at sign-in and blocks devices that fall short.
If you allow personal devices, split the access model. Managed devices get full access. Unmanaged devices can be limited to browser-only sessions, using app enforced restrictions or Defender for Cloud Apps session controls where your applications and licences support them. These controls can block downloads, and in some apps copy and print, but they do not stop screenshots or a photo of the screen.
Network segmentation in practice
Start with your highest-value assets: finance systems, domain controllers, backup infrastructure and any operational technology (OT). Isolate them with deny-by-default inbound rules that permit only the flows each service needs. You do not have to segment the whole network to get most of the value.
Zero Trust Network Access (ZTNA) services such as Cloudflare Access, Microsoft Entra Private Access and Zscaler Private Access sit in front of internal applications and check identity and device before opening a connection. A traditional VPN usually gives network-level access once connected, limited only by whatever segmentation and filtering you have added. ZTNA brokers access per application, which limits what a stolen credential can reach. It still depends on well-written policies, and some ZTNA products can also be configured for broader network access.
Zero Trust and regulatory compliance
None of the main frameworks requires Zero Trust by name. Zero Trust contributes to the access control, authentication, least privilege and network security controls they ask for. Those frameworks also require risk management, incident handling, business continuity, supplier security, documentation and evidence that the controls work. Implementing Zero Trust supports those requirements; it does not show conformity on its own.
UK NIS Regulations 2018. Operators of essential services and relevant digital service providers are often assessed against the NCSC's Cyber Assessment Framework, whose principle B2 (identity and access control) maps closely to the identity and device pillars. The Cyber Security and Resilience Bill, still before Parliament, would extend the regime to more organisations, including managed service providers. The NCSC also publishes its own zero trust architecture design principles.
NIS2. If you provide services in the EU, NIS2 may apply to that activity. Article 21(2)(i) covers access control policies and Article 21(2)(j) covers multi-factor or continuous authentication.
DORA. The Digital Operational Resilience Act (Regulation (EU) 2022/2554) has applied to EU financial entities since 17 January 2025. Its ICT risk management rules, detailed in Delegated Regulation (EU) 2024/1774, include network segmentation, identity management, access control and strong authentication. UK firms feel it mainly through EU subsidiaries or as ICT providers to EU financial entities. See our DORA gap analysis guide.
ISO 27001:2022 Annex A. The controls below map most directly to a Zero Trust programme:
| Control | Zero Trust practice | Evidence an auditor may ask for |
|---|---|---|
| A.5.15 Access control | Policy-based access per application | Access control policy, Conditional Access export |
| A.5.16 Identity management | Lifecycle for user and non-human identities | Joiner, mover and leaver records |
| A.5.17 Authentication information | MFA, passwordless, secret rotation | MFA registration report, secrets policy |
| A.5.18 Access rights | Least privilege, periodic reviews | Access review results |
| A.8.2 Privileged access rights | Just-in-time admin roles (PIM) | Role activation logs |
| A.8.3 Information access restriction | Per-resource authorisation, labels | Permission reports, label policy |
| A.8.5 Secure authentication | Phishing-resistant MFA, legacy auth blocked | Sign-in logs, policy settings |
| A.8.20 Networks security | Encrypted traffic, deny by default | Firewall and ZTNA rule sets |
| A.8.21 Security of network services | ZTNA and DNS filtering requirements | Service agreements, configuration |
| A.8.22 Segregation of networks | Segmented zones for critical assets | Network diagrams, segmentation tests |
You still need a risk assessment, a Statement of Applicability and records that the controls operate. See the ISO 27001 guide for SMEs.
What Zero Trust does not solve on its own
Zero Trust is not a certification, a product or a substitute for a compliance programme. A first deployment usually covers staff identities, devices and remote access. Plan for these gaps as well:
- Workload identities. Service principals, managed identities, API keys and other secrets need an owner, narrowly scoped permissions and rotation, just like user accounts.
- Phishing-resistant MFA. Push approvals and SMS codes can be phished or abused through MFA fatigue. Move administrators and other high-risk access to FIDO2 security keys, passkeys or Windows Hello for Business first.
- Legacy applications. An application without modern authentication needs an application proxy, a ZTNA connector or its own isolated segment until it is replaced.
- Policy lifecycle. Test new Conditional Access policies in report-only mode, record every exception with an owner and an end date, and review policies when staff, applications or licences change.
- Identity provider outages. Keep at least two cloud-only emergency access (break-glass) accounts that cannot be locked out by your normal policies. Protect them with strong authentication, alert on every sign-in and test them on a schedule.
- Vulnerabilities and detection. Identity controls do not stop the exploitation of an unpatched internet-facing system. You still need vulnerability management, logging that covers the systems that matter and detections that are tested.
How to implement Zero Trust in practice
Work area by area, starting where your risk assessment points. Each row pairs a first control with a way to check it works.
| Area | Initial control | How to verify |
|---|---|---|
| Identity | Audit all user and service accounts, enforce MFA, block legacy authentication | MFA coverage report; no legacy sign-ins in the last 30 days |
| Privileges | Remove standing admin rights, use time-limited activation | Count of permanent admin assignments; activation logs |
| Devices | Manage corporate devices and require compliance for sensitive apps | Compliance dashboard; blocked sign-ins from non-compliant devices |
| Network | Segment critical assets and replace broad VPN access where possible | Rule review; test that a user subnet cannot reach backup or domain controllers |
| Applications | Put internal apps behind authenticated access, remove anonymous access | Application inventory with authentication method for each |
| Data | Label sensitive data and add DLP for the riskiest transfers | Label coverage; DLP matches reviewed each month |
| Monitoring | Send identity and access logs to a SIEM with alerts on risky events | Alerts for impossible travel, new device sign-ins and out-of-hours role activation fire in testing |
Zero Trust assumes a breach can happen. The design question is how much an attacker can reach from their first foothold.
Zero Trust architecture is one workstream inside a security programme, not the whole programme. A cybersecurity roadmap places it alongside identity governance, endpoint management and incident response so the controls build on each other.
Ryland has delivered cybersecurity, compliance, and IT management programmes for regulated organisations across the UK and the Netherlands for over 20 years. View full profile