Skip to content

Zero Trust Network Architecture

Zero Trust Network Architecture (ZTNA) eliminates implicit trust from every network interaction. Unlike traditional perimeter-based security — where anything inside the firewall is trusted — ZTNA treats every access request as if it originates from an untrusted network. This document provides a vendor-neutral reference architecture aligned with NIST SP 800-207 and the BeyondCorp model.


Core Principles

Zero Trust is built on three non-negotiable principles:

Principle Definition Network Implication
Never trust, always verify Every access request is fully authenticated, authorized, and encrypted regardless of origin No network location grants implicit trust; internal traffic is treated identically to external
Least privilege Grant minimum access required for the specific task at the specific time Microsegmented access policies replace broad network ACLs; no flat network access
Assume breach Design controls assuming the adversary is already inside the perimeter Lateral movement is blocked by default; east-west traffic is inspected and logged

Traditional perimeter security assumes that once a user or device passes the firewall, it can be trusted. This assumption fails catastrophically when attackers gain initial access through phishing, supply chain compromise, or credential theft — they inherit the implicit trust of the internal network.


Zero Trust Network Reference Architecture

flowchart TB
    subgraph Users["Users & Devices"]
        EU[Enterprise User]
        RU[Remote User]
        SVC[Service Account]
        IOT[IoT / OT Device]
    end

    subgraph DeviceTrust["Device Trust Layer"]
        DA[Device Agent]
        DC[Device Compliance Check]
        CERT[Certificate Validation]
    end

    subgraph ControlPlane["Control Plane"]
        PAP[Policy Administration Point<br/>Policy authoring & lifecycle]
        PDP[Policy Decision Point<br/>Real-time access decisions]
        PIP[Policy Information Point<br/>Context aggregation]
    end

    subgraph DataSources["Context Sources → PIP"]
        IDM[Identity Provider / IdP]
        TI[Threat Intelligence Feed]
        SIEM_CTX[SIEM Risk Scores]
        CMDB[Asset Inventory / CMDB]
        UEBA[UEBA Behavioral Baseline]
    end

    subgraph DataPlane["Data Plane"]
        PEP[Policy Enforcement Point<br/>Inline gateway / proxy]
        MICRO[Microsegmentation Engine]
        ENCRYPT[mTLS / Encrypted Tunnels]
    end

    subgraph Resources["Protected Resources"]
        APP[Applications]
        DATA[Data Stores]
        INFRA[Infrastructure Services]
        CLOUD[Cloud Workloads]
    end

    EU --> DA
    RU --> DA
    SVC --> DA
    IOT --> DA
    DA --> DC
    DC --> CERT
    CERT --> PEP

    PEP -- "access request" --> PDP
    PDP -- "query context" --> PIP
    PIP --> IDM
    PIP --> TI
    PIP --> SIEM_CTX
    PIP --> CMDB
    PIP --> UEBA
    PAP -- "publish policy" --> PDP
    PDP -- "allow/deny" --> PEP

    PEP --> MICRO
    MICRO --> ENCRYPT
    ENCRYPT --> APP
    ENCRYPT --> DATA
    ENCRYPT --> INFRA
    ENCRYPT --> CLOUD

Core Components

The four pillars of a Zero Trust architecture, derived from NIST SP 800-207:

Policy Decision Point (PDP)

The PDP is the brain of the architecture. It receives access requests from enforcement points, evaluates them against policy, and returns an allow/deny decision.

Attribute Detail
Function Evaluate access request against policy using contextual signals
Inputs User identity, device posture, resource sensitivity, behavioral risk, threat intelligence
Output Allow, deny, or step-up authentication decision
Latency target < 50ms per decision (P99)
Availability 99.99% — failure mode must default to deny

The PDP should never be a single point of failure. Deploy in an active-active configuration across availability zones. A PDP outage that defaults to 'allow' defeats the purpose of Zero Trust.

Policy Enforcement Point (PEP)

The PEP sits inline with every access request and enforces the PDP's decision. It is the only path to protected resources.

PEP deployment patterns:

  • Identity-aware proxy — Reverse proxy authenticating every HTTP/S request (BeyondCorp model)
  • Software-defined perimeter (SDP) — Single-packet authorization before connection establishment
  • Microsegmentation agent — Host-based firewall enforcing workload-to-workload policy
  • API gateway — Token validation and scope enforcement for service-to-service calls

Policy Information Point (PIP)

The PIP aggregates contextual signals that feed PDP decisions:

Signal Source Data Provided Update Frequency
Identity Provider (IdP) User identity, group membership, authentication strength Real-time
Device Management (MDM/EDR) OS version, patch status, encryption state, malware status Every 5 minutes
Threat Intelligence Known malicious IPs, IOCs, campaign context Every 15 minutes
SIEM / UEBA User risk score, anomaly detection results Near real-time
CMDB / Asset Inventory Resource classification, owner, data sensitivity Daily sync
Geolocation / Network Source IP, location, impossible travel detection Per request

Policy Administration Point (PAP)

The PAP is where policies are authored, tested, versioned, and published to PDPs.

Policy lifecycle:

Author policy → Peer review → Test in simulation mode → Staged rollout (5% → 25% → 100%) → Monitor → Iterate

Treat Zero Trust policies as code. Store them in version control, require peer review for changes, test in simulation mode before enforcement, and maintain rollback capability for every policy change.


Zero Trust Maturity Model

Organizations progress through maturity levels as they adopt Zero Trust. This model provides a self-assessment framework:

Level Name Identity Network Devices Applications Data Automation
L0 Traditional Perimeter Passwords only; VPN = trusted Flat internal network; perimeter firewall only No device compliance checks No app-level auth beyond network access No data classification Manual firewall rules
L1 Basic Segmentation MFA for external access VLANs for major zones (DMZ, internal, servers) Corporate-managed devices required Basic SSO for major apps Sensitivity labels on some data Scripted ACL management
L2 Identity-Aware MFA everywhere; SSO for all apps Identity-aware proxy for web apps Device compliance checked at login Per-app authorization policies Data classification for regulated data Policy-as-code for identity
L3 Microsegmented Risk-based adaptive authentication Microsegmentation for workloads; east-west filtering Continuous device posture assessment Per-request authorization; no VPN DLP policies enforced at access point Automated policy deployment
L4 Continuous Verification Continuous authentication; session risk scoring Encrypted east-west; no implicit trust zones Real-time device risk scoring; auto-quarantine Just-in-time app access; session recording Real-time data access auditing Automated incident response for policy violations
L5 Full Zero Trust Passwordless; behavioral biometrics Full SDP; network location irrelevant Zero-touch provisioning; hardware attestation Every request verified; all sessions ephemeral Automated data governance; field-level encryption Self-healing policy; autonomous threat response

Maturity Assessment

Most organizations begin at L0–L1. A realistic 24-month goal is reaching L3 across identity and network pillars, with L2 in devices and data. Full L5 maturity typically requires 3–5 years of sustained investment.


Implementation Roadmap

A six-phase adoption plan aligned with the Forrester ZTX and NIST frameworks:

Phase 1: Identify Protect Surfaces (Months 1–2)

Define what you are protecting. The protect surface is the inverse of the attack surface — it is small, well-defined, and defensible.

Protect Surface Category Examples Discovery Method
Critical data (DAAS) Customer PII, financial records, IP, credentials Data classification scan; DLP inventory
Critical applications ERP, CRM, SIEM, CI/CD pipeline Application portfolio review
Critical assets Domain controllers, certificate authorities, key management CMDB and infrastructure audit
Critical services DNS, DHCP, Active Directory, authentication services Service dependency mapping

Deliverable: Prioritized register of protect surfaces ranked by business impact.

Phase 2: Map Transaction Flows (Months 2–3)

Understand how traffic flows to and from each protect surface. You cannot protect what you do not understand.

flowchart LR
    subgraph Sources["Traffic Sources"]
        ANALYST[Analysts]
        APPS[Applications]
        SVCS[Services]
        PARTNERS[Partners]
    end

    subgraph Flows["Transaction Flows"]
        AUTH[Authentication Flow]
        API_FLOW[API Call Flow]
        DATA_FLOW[Data Access Flow]
        ADMIN_FLOW[Admin Flow]
    end

    subgraph Surfaces["Protect Surfaces"]
        CRM_PS[CRM — Customer Data]
        SIEM_PS[SIEM — Security Logs]
        CICD_PS[CI/CD — Source Code]
        AD_PS[AD — Identity Infrastructure]
    end

    ANALYST --> AUTH --> CRM_PS
    APPS --> API_FLOW --> SIEM_PS
    SVCS --> DATA_FLOW --> CICD_PS
    PARTNERS --> ADMIN_FLOW --> AD_PS

Deliverable: Transaction flow map showing all access paths to each protect surface, including ports, protocols, and identity context.

Phase 3: Build Zero Trust Architecture (Months 3–6)

Deploy the core components (PDP, PEP, PIP, PAP) starting with the highest-priority protect surfaces.

Priority order:

  1. Identity infrastructure (IdP, AD, certificate services)
  2. Security tooling (SIEM, SOAR, EDR consoles)
  3. Developer infrastructure (CI/CD, source control, artifact repositories)
  4. Business-critical applications (ERP, CRM)
  5. General enterprise applications
  6. IoT and OT environments

Do not attempt to deploy Zero Trust everywhere simultaneously. Start with one protect surface, validate the architecture, then expand. A failed big-bang rollout creates security gaps worse than the legacy perimeter.

Phase 4: Create Zero Trust Policy (Months 5–8)

Define granular access policies using the Kipling Method — who, what, when, where, why, and how:

Policy Element Question Example Policy Statement
Who Which identity? Only members of sg-soc-analysts with active MFA session
What Which resource? SIEM query API — read-only scope
When What time window? Business hours (06:00–22:00 local) for standard access
Where From which location/device? Corporate-managed device with compliant posture score ≥ 80
Why What is the business justification? Active incident investigation (ticket reference required for sensitive data)
How Through which access path? Via identity-aware proxy; direct network access denied

Phase 5: Monitor and Maintain (Months 8–12)

Deploy continuous monitoring to validate Zero Trust controls:

  • Policy violation monitoring — Alert on every denied access attempt; investigate repeated denials
  • Configuration drift detection — Automated checks that PEP configurations match PAP-published policy
  • Access pattern analytics — Baseline normal access patterns; alert on deviations (UEBA)
  • Control effectiveness metrics — Track the metrics defined in the Metrics section below

Phase 6: Automate and Orchestrate (Months 10–18)

Mature from manual policy management to automated orchestration:

Automation Target Manual State Automated State
Policy updates Admin manually edits firewall rules Policy-as-code deployed via CI/CD pipeline
Incident response Analyst manually blocks compromised account SOAR playbook auto-disables account and triggers step-up auth
Device quarantine Help desk manually removes device from network EDR auto-quarantine triggers PDP policy update within 30s
Access reviews Quarterly spreadsheet review Continuous access certification with auto-revocation of unused permissions
Threat response Manual IOC blocking across multiple tools Threat intelligence auto-propagates to PDP deny lists within 5 minutes

Technology Components

Software-Defined Perimeter (SDP)

SDP implements a "dark cloud" — protected resources are invisible to unauthorized users. Connection is established only after authentication and authorization.

1. SDP Controller authenticates user + device
2. Controller validates against PDP
3. Controller provisions encrypted tunnel to specific resource
4. Resource is never exposed to unauthenticated scanning

SDP follows a 'authenticate first, connect second' model — the opposite of traditional networking where connection is established first and authentication happens at the application layer.

Zero Trust Network Access (ZTNA)

ZTNA replaces VPN by providing per-application access rather than full network access:

Capability Legacy VPN ZTNA
Access granularity Full network layer access Per-application, per-session access
Device trust Optional or basic check Continuous posture validation
Lateral movement risk High — user has network access Minimal — no network layer access granted
User experience Client software + split-tunnel complexity Seamless browser or lightweight agent
Scalability VPN concentrator bottleneck Cloud-delivered, horizontally scalable
Visibility Limited to connection logs Full request-level logging with identity context

Microsegmentation

Microsegmentation enforces security policy at the individual workload level rather than the network perimeter:

Implementation approaches:

  • Agent-based — Host-level firewall managed by central policy (most granular)
  • Network-based — SDN controller programs switches/routers (infrastructure-dependent)
  • Hypervisor-based — Virtual firewall between VMs (virtualized environments only)
  • Cloud-native — Security groups, network policies (Kubernetes NetworkPolicy, AWS SGs)

Identity-Aware Proxy (IAP)

IAP authenticates and authorizes every request at the proxy layer before forwarding to the backend application:

sequenceDiagram
    participant U as User
    participant IAP as Identity-Aware Proxy
    participant IDP as Identity Provider
    participant PDP as Policy Decision Point
    participant APP as Application

    U->>IAP: HTTPS request to app.example.com
    IAP->>IDP: Validate session / redirect to auth
    IDP-->>IAP: Identity token (user, groups, MFA status)
    IAP->>PDP: Authorize (identity + device + context)
    PDP-->>IAP: Allow (scoped permissions)
    IAP->>APP: Forward request with identity headers
    APP-->>IAP: Response
    IAP-->>U: Response (logged with full context)

Secure Access Service Edge (SASE)

SASE converges networking and security into a cloud-delivered service:

SASE Component Function Zero Trust Role
ZTNA Per-app access Replaces VPN; enforces least privilege
SWG (Secure Web Gateway) Web traffic inspection Inline threat detection for outbound traffic
CASB Cloud app control Visibility and policy enforcement for SaaS
FWaaS Cloud firewall Network-layer policy for branch and remote
SD-WAN WAN optimization Secure transport; traffic steering to SASE PoPs

Vendor-Neutral Reference Architecture

This architecture is aligned with NIST SP 800-207 and the BeyondCorp model. It does not depend on any single vendor.

NIST SP 800-207 Deployment Models

Model Description Best For
Enhanced Identity Governance Access decisions based primarily on identity and attributes Organizations with mature IAM; SaaS-heavy environments
Micro-segmentation Network-level enforcement at the workload level Data center and cloud IaaS workloads
Network Infrastructure + SDP Network overlays and SDP controllers mediate all access Organizations with complex hybrid infrastructure

Most real-world deployments combine all three models. Use identity governance for user-to-application access, microsegmentation for workload-to-workload, and SDP for infrastructure access.

BeyondCorp Reference

Google's BeyondCorp model demonstrates that enterprise applications can be accessed securely without a VPN:

  1. Access depends on device and user credentials — not network location
  2. All access to enterprise resources is authenticated, authorized, and encrypted
  3. Access is granted on a per-request basis — no persistent trust
  4. Policy is dynamic and sourced from multiple contextual signals

Integration Points

Zero Trust architecture does not exist in isolation. It integrates with the broader security operations stack:

Integration Direction Data Exchanged Purpose
SIEM PDP → SIEM Access decisions, policy violations, denied requests Centralized logging and correlation of all access events
SIEM SIEM → PIP User and entity risk scores Dynamic policy adjustment based on SIEM analytics
SOAR PDP → SOAR High-severity policy violations Automated incident response for Zero Trust violations
SOAR SOAR → PDP Block/allow directives Automated policy updates during incident response
EDR EDR → PIP Device health, malware detection, IOCs Real-time device posture feeding access decisions
NDR NDR → PIP Anomalous network behavior, lateral movement detection Network-level context for access decisions
CASB PEP ↔ CASB SaaS access policy and enforcement Unified policy for on-prem and cloud applications
IAM IAM ↔ PDP Identity attributes, group membership, auth strength Core identity context for every access decision
flowchart LR
    PDP[Policy Decision Point]

    SIEM[SIEM] <-->|risk scores ↔ access logs| PDP
    SOAR[SOAR] <-->|incidents ↔ policy updates| PDP
    EDR[EDR] -->|device posture| PDP
    NDR[NDR] -->|network anomalies| PDP
    CASB[CASB] <-->|SaaS policy| PDP
    IAM[IAM] <-->|identity context| PDP

Metrics and KPIs

Track these metrics to measure Zero Trust effectiveness and maturity:

Coverage Metrics

Metric Definition Target (L3 Maturity) Target (L5 Maturity)
% of resources behind ZT controls Protect surfaces with PDP/PEP enforcement ≥ 80% 100%
% of access requests evaluated by PDP Requests going through policy evaluation vs. bypassing ≥ 90% 100%
% of users on passwordless / phishing-resistant MFA Users with hardware key or biometric auth ≥ 60% ≥ 95%
% of workloads microsegmented Workloads with granular east-west policy ≥ 50% ≥ 95%

Effectiveness Metrics

Metric Definition Target
Policy violation rate Denied access requests / total requests (high rate may indicate misconfiguration) < 5% sustained (after tuning)
Lateral movement detection rate % of simulated lateral movement attempts detected and blocked ≥ 95%
Mean time to policy update Time from threat detection to PDP policy enforcement < 15 minutes (automated)
False positive rate (policy denials) Legitimate access incorrectly blocked < 1% of total access requests
PDP decision latency (P99) 99th percentile latency for access decisions < 100ms

Operational Metrics

Metric Definition Target
PDP availability Uptime of policy decision infrastructure ≥ 99.99%
Policy drift incidents Instances where PEP config diverges from PAP-published policy 0 per month
Credential exposure events Service account credentials found outside vault 0 per quarter
Access review completion rate % of access reviews completed on schedule 100%

A Zero Trust architecture is only as strong as its monitoring. If policy violations are not tracked, if PDP decisions are not logged, or if metrics are not reviewed — the architecture provides a false sense of security.


Common Zero Trust Network Mistakes

Avoid These

  • Treating Zero Trust as a product purchase. Zero Trust is an architecture and strategy, not a single vendor solution. No product delivers "Zero Trust in a box."
  • Deploying ZTNA but keeping VPN as a fallback. If VPN remains available, users and attackers will use it. The legacy path must be decommissioned.
  • Flat microsegmentation policies. Microsegmentation with overly permissive policies (allow-any within segment) provides no value. Policy must be granular.
  • Ignoring east-west traffic. Most Zero Trust projects focus on north-south (user to app). Lateral movement between workloads is the actual kill chain vector.
  • PDP single point of failure. If the PDP fails and defaults to "allow," the entire architecture collapses. Failure mode MUST be deny.
  • No continuous monitoring. Zero Trust without continuous verification is just a one-time gate check — not Zero Trust.

Nexus SecOps Control Mapping

Zero Trust Network Control Nexus SecOps Controls
Microsegmentation enforcement Nexus SecOps-121, Nexus SecOps-122
Identity-aware proxy deployment Nexus SecOps-111, Nexus SecOps-113
PDP policy management Nexus SecOps-114, Nexus SecOps-205
Device trust validation Nexus SecOps-130, Nexus SecOps-131
Continuous access monitoring Nexus SecOps-049, Nexus SecOps-119
Credential vault integration Nexus SecOps-104, Nexus SecOps-112
Lateral movement detection Nexus SecOps-062, Nexus SecOps-065
Network traffic encryption Nexus SecOps-123, Nexus SecOps-125
Access decision logging Nexus SecOps-005, Nexus SecOps-007

See Zero Trust SOC | Reference Architecture | Integration Patterns Related chapters: Chapter 13: Governance | Chapter 33: Identity & Access Security | Chapter 22: Network Security