What is a zero-trust architecture and how does it differ from perimeter security?

Traditional perimeter security works like a castle and moat: once you're inside the network (past the firewall/VPN), you're implicitly trusted to access most internal systems. This model fails badly when an attacker gets inside — via a compromised endpoint, a phished employee, or a misconfigured cloud service — because they then move laterally freely.

Zero Trust operates on the principle 'never trust, always verify.' Every request — regardless of whether it originates inside or outside the corporate network — must be authenticated, authorised against the specific resource, and encrypted. The pillars are: strong per-user and per-service identity; device posture verification (is this a managed, up-to-date device?); least-privilege authorisation re-evaluated on every request; microsegmentation so a compromised service can't reach unrelated services; mutual TLS (mTLS) for service-to-service communication so both sides prove their identity; and pervasive logging.

In practice: identity-aware access proxies (Google's BeyondCorp model) replace blanket VPN access; service meshes (Istio, Linkerd) enforce mTLS between microservices instead of relying on flat network trust.