Zero Trust and Secure Architecture on Azure: Defense in Depth, Landing Zone Security, Threat Modeling and Review Checklist
Zero Trust Principles
| Principle | Meaning | Azure examples |
|---|---|---|
| Verify explicitly | Authenticate and authorize using all signals | MFA, Conditional Access, device compliance |
| Use least privilege | Just enough and just-in-time access | RBAC, PIM, JIT VM access, managed identities |
| Assume breach | Limit blast radius, monitor and encrypt | Segmentation, Defender, Sentinel, encryption |
Defense in Depth: Layers
| Layer | Controls |
|---|---|
| Physical | Microsoft datacenter security |
| Identity | Entra ID, MFA, Conditional Access, PIM, passwordless |
| Perimeter | DDoS Protection, Front Door + WAF, Azure Firewall |
| Network | Segmentation, NSGs, private endpoints, no public IPs, Bastion |
| Compute | Patching, Defender for Servers, trusted launch, JIT |
| Application | Secure coding, secrets in Key Vault, WAF, dependency scanning |
| Data | Encryption, classification, Purview, backups, immutability |
| Operations | Monitor, Sentinel, Policy, incident response |
Shared Responsibility
| On-prem | IaaS | PaaS | SaaS | |
|---|---|---|---|---|
| Data and identities | You | You | You | You |
| Apps | You | You | Shared | Microsoft |
| OS | You | You | Microsoft | Microsoft |
| Network controls | You | Shared | Shared | Microsoft |
| Physical | You | Microsoft | Microsoft | Microsoft |
You are always responsible for your data, identities and access.
Threat Modeling with STRIDE
| Threat | Question | Typical control |
|---|---|---|
| Spoofing | Can someone pretend to be a user or service? | Strong auth, managed identity |
| Tampering | Can data be changed? | Signing, integrity checks, immutability |
| Repudiation | Can an action be denied? | Audit logs |
| Information disclosure | Can data leak? | Encryption, private endpoints, RBAC |
| Denial of service | Can it be overloaded? | DDoS, autoscale, rate limits |
| Elevation of privilege | Can a user gain more rights? | Least privilege, PIM |
Secure Reference Design
- Users sign in with Entra ID + MFA + Conditional Access and reach the app through Front Door with WAF
- The app runs in a spoke VNet, with no public IPs on VMs. Admins use Bastion and PIM
- The app accesses SQL, Storage and Key Vault with managed identities over private endpoints
- All traffic to the internet goes through Azure Firewall in the hub
- Policy enforces tags, allowed regions, encryption and diagnostics
- Logs flow to Log Analytics and Sentinel. Defender for Cloud monitors posture
- Backups are immutable and replicated. IaC and pipelines use OIDC with least privilege
Architecture Review Checklist
- Is MFA enforced for everyone, and are break-glass accounts protected and monitored?
- Are standing Owner/Contributor assignments minimized and replaced with PIM?
- Do any resources have public endpoints? Can private endpoints replace them?
- Are secrets in Key Vault and managed identities used instead of passwords?
- Is data encrypted at rest and in transit, and who controls the keys?
- Are NSG/firewall rules specific, with no any-any rules and no open management ports?
- Is logging enabled, centralized, retained and alerting?
- Are backups tested, protected from deletion and meeting RPO/RTO?
- Do Policy and Defender show non-compliance, and who fixes it?
- Is there an incident response plan and regular exercises?
Practice
- Run STRIDE on a simple web app with a database and write one control per threat
- Review a sample design with the checklist and list five findings
- Explain Zero Trust to a manager in three sentences
Conclusion
Secure designs combine identity, network, compute, application, data and operations controls following Zero Trust: verify explicitly, use least privilege and assume breach. Use STRIDE and a checklist during every design review.
Quick Summary: Zero Trust; layers; shared responsibility; STRIDE; reference design; checklist.

