Zero trust
A security model that assumes no network is trusted; every request is authenticated and authorized regardless of source. Replaces perimeter security. Implemented with mTLS, identity, and per-request policy.
- Cloud and hybrid environments.
- Any organization moving beyond perimeter security.
- Isolated, fully trusted legacy networks (rare).
- Without the PKI to support it.
Definition
Zero trust means 'never trust, always verify': no implicit trust based on network location. Every connection authenticates (often mTLS), authorizes per request, and is encrypted. Identity, not network, grants access.
Why it exists
Perimeter security fails once an attacker is inside. Cloud and remote work dissolved the perimeter. Zero trust treats every call as untrusted, limiting lateral movement.
How it works
Issue short-lived identities to workloads (SPIFFE) and users (OIDC). Enforce mTLS on every connection. A policy engine authorizes each request by identity + context. Microsegmentation limits blast radius. Continuous verification replaces one-time auth.
Scaling limits
Every request pays auth/encryption overhead. Identity issuance, rotation, and policy at scale need automation. Legacy apps may not support mTLS without a sidecar.
Cost shape
PKI/identity infra + per-request authz overhead + ops. Blow-ups: misconfigured policies and over-broad access.
Alternatives
- Perimeter (VPN/firewall) — Simpler but fails once breached.
- Network ACLs — Coarse network-level trust, no identity.
Failure modes
- Over-broad policies negating the model.
- Identity issuance failures blocking all traffic.
- Legacy gaps.