Microsoft’s retirement list records AZ-500 Azure Security Engineer Associate as retired on August 31, 2026. Azure security design remains current work, but readers should not book AZ-500 on the assumption that its historical exam guide is still active. This article examines the design responsibilities behind effective identity, network, data and workload protection, distinct from a set of short exam drills.
Define the security boundary before adding controls
Start with assets, attack paths, regulatory obligations and likely operators. Separate the identity perimeter from subscriptions, virtual networks, data stores and application-level permissions. A perimeter firewall will not compensate for an application identity that can read every storage account in the tenant. Build a threat model that maps privileged actions to the resources they can affect.
For a multi-subscription deployment, document who can create role assignments and which management-group policies prevent deviations. Ownership must be explicit where production and security teams share responsibilities.
Design privileged access around specific operations
Interactive administrators, deployed workloads and automated release pipelines should not share the same broad identity. Use Entra roles for directory-level responsibilities and Azure RBAC for resource scopes, understanding where the policies intersect. Prefer managed identities where supported; protect any unavoidable secrets in a managed vault with tested rotation.
A user who cannot read a Key Vault secret might be missing an effective data-plane permission, not the ability to see the vault resource in the portal. Check the correct control plane before expanding privileges.
Segment networks by flows, not arbitrary address ranges
For a cross-subscription application, the design might include an internet-facing gateway, internal APIs, a managed database and key-management services. Start from data flows and identify the identities and allowed network paths at each hop. Verify whether a compromised application identity could enumerate production subscriptions or export customer data. An isolated virtual network will not prevent that if the identity has broad control-plane privileges. Add scoped roles and organization-level policies, then test authorized deployment and prohibited escalation in separate accounts. Record failure evidence at each layer; otherwise a security diagram may show intended isolation while the actual policies grant much broader reach than the architecture requires.
Network security groups, firewalls, private endpoints and application gateways provide different policy decisions. Model an actual client-to-service flow, including DNS lookup, routing, TLS termination and outbound dependencies. A private endpoint changes how a service is reached; it does not automatically make every consuming application secure.
For a data service with public access disabled, test both approved private clients and an external client that must fail. Confirm the observed route and authorization rather than relying on a portal toggle alone.
Apply data protection through the application lifecycle
Encryption at rest and TLS in transit matter, but data may also leak through logs, exports, backups or broadly accessible development environments. Classify datasets, define key and secret ownership, and decide how test teams receive safely minimized data. Customer-managed keys add control but can also introduce availability risk if key access is revoked unexpectedly.
Design recovery for lost keys or corrupted configuration without assuming protected data can always be recovered. Measure backup success by a restore and correct decryption, not by backup-job status.
Make detection and response operational
Architecture reviews should also distinguish preventive controls from detective controls. An Azure Policy assignment might prevent a public storage setting in one subscription; an activity log alert might report the same setting after it changes elsewhere. One cannot replace the other. Map each sensitive resource to the preventive restriction, evidence source, incident owner and tested response. When the team discovers a gap, decide whether the safest fix is a policy, narrower identity scope, application change or improved monitoring. Evaluate developer usability and emergency exception behavior before applying organizational rules at scale. This connects security governance to the practical likelihood that teams will operate within the controls.
Defender for Cloud recommendations and security logs can help identify risk, but an alert has value only when the team knows what to do. Define triage inputs, asset criticality, evidence retention and containment decisions. A high-severity finding on an isolated test resource may be lower priority than a privilege escalation affecting production identity.
In an incident rehearsal, preserve activity logs before changing access controls. Revoke compromised sessions and credentials in a measured order, then confirm application access and event completeness afterward.
Plan governance that developers can obey
Incident recovery has architectural consequences, too. Suppose a workload’s storage encryption relies on a customer-managed key whose permission was accidentally revoked during an urgent security response. Monitoring might report both storage access errors and application unavailability. The recovery plan must distinguish restoring legitimate decrypt permissions from weakening the security boundary permanently. Identify who can approve a key-policy change, how application traffic is drained and how data integrity is checked after recovery. Document and rehearse the path in a nonproduction environment. A secure design balances confidentiality, integrity and availability; switching every security option on without examining its failure modes is not sufficient engineering judgment.
Guardrails such as Azure Policy, resource locks and approved deployment modules can make secure defaults repeatable. Overly rigid policies can also delay urgent repairs and drive risky workarounds. Describe allowed exceptions, required approvals and review timeframes. Test policy assignments in a nonproduction scope before applying them to many subscriptions.
The historical AZ-500 domains help organize security topics, but modern learning should follow active product documentation and the issuer’s current credentials.
Secure Azure architecture is a set of accountable decisions about access, network paths, sensitive data and recoverability. The value of those decisions is independent of the retired AZ-500 credential.