Zero Trust Network Architecture

No Change
adopt
First Added:October 1, 2024 Updated: July 2, 2026

Zero Trust Network Architecture. (zero trust) assumes no implicit trust based on network location.

Summary

Core idea: without a trusted “inside” network, you must enforce safety on every request. Perimeter firewalls and VPNs that grant broad LAN access do not meet that bar.

Common pillars (NIST-style):

PillarPractice
Verify explicitlyAuthN/AuthZ on every request; device and user posture
Least privilegeJust enough access for the task; time-bound where possible
Assume breachSegment, log, and monitor; no silent lateral movement

How we apply it: treat ZTNA as the default remote-access pattern. Pair Access on Demand for production and break-glass paths. Use Tailscale or similar L3 VPN only when an app truly needs network adjacency, not as the org-wide front door.

Not the same as: ZTNA (product and broker pattern) or a single vendor SKU. This note is the reference architecture under Specification.

Details

  • Origin: popularized in industry by Google’s BeyondCorp papers (see Garden Notes). NIST SP 800-207 documents a widely cited zero-trust architecture model.
  • Identity: Single Sign-on (OIDC/SAML) is the front door; group and role claims feed Policy as Code and access brokers.
  • Products (examples): Boundary (Hashicorp) for session-brokered infra access; commercial ZTNA suites when you need a managed broker fleet.
  • Contrast: “castle and moat” VPNs, flat VPC peering, and long-lived SSH keys on bastions are hold patterns for new work.