Zero trust

In one line

Zero trust abandons the castle-and-moat: being inside the hospital network grants nothing — every request is authenticated, authorised against policy, and encrypted, every time.

The problem it solves

The old model trusted the network: get inside the firewall and you were assumed friendly. That assumption is dead — phishing, stolen credentials, and supply-chain compromise put attackers inside routinely, and a flat internal network lets them roam. Zero trust starts from the opposite premise: assume breach, and verify every request as if it came from the open internet.

The principles (NIST 800-207)

• Strong identity for users and workloads — every actor proves who it is, every time.
• Device posture checks — is the endpoint patched, managed, healthy?
• Least privilege per request — grant the minimum needed for this action, not standing broad access.
• Micro-segmentation — divide the network so a compromised workstation can't reach everything.
• Continuous verification — no single login that unlocks the day; re-evaluate on context.
• Rich telemetry — log and analyse, because you assume something will get through.

It's a journey, not a product

You can't buy "zero trust" in a box. It's an architecture journey: identity first, then segmentation, then policy automation. Vendors sell pieces; the model is how you assemble them.

Where it shows up in digital health

Hospitals are flat-network heaven for ransomware — one phished laptop reaching every unsegmented device is the recurring incident, and it shuts down emergency departments. Zero trust is the counter-design:

• Medical devices on tightly segmented zones.
• EHR access gated per request.
• Vendor remote access brokered, not VPN'd into the flat network.

It connects to HIPAA's Security Rule, but goes well beyond the compliance floor. And it's the same philosophy this platform uses at the database layer: Supabase row-level security grants nothing by default — every row access is policy-checked.

Common pitfalls

• "We bought a zero-trust product" — it's an architecture, not a SKU.
• Identity without segmentation — strong login but a flat network still lets attackers roam.
• Friction that drives workarounds — clinicians will defeat security that blocks care; design for the workflow.

Key takeaways

• Zero trust: never trust the network, always verify the request — assume breach.
• Identity + device posture + least privilege + micro-segmentation + continuous verification.
• It's a staged architecture journey, not a purchase.
• In health it's the antidote to flat-network ransomware — and the model behind RLS.