Cloud and AI Security Engineer Flashcards & Memory Notes

Just-in-time elevation of privileged Azure or Entra roles for a limited window, often with approval and MFA. It is not a standing Owner assignment.

No. All applicable enabled policies are evaluated, and an applicable block prevents access. Test policies with report-only evaluation and sign-in evidence; there is no administrator-defined first-match ordering.

In Azure Key Vault, retrieved at runtime—preferably with managed identity. Do not bake secrets into images or App Configuration.

Yes. It can hold a reference while the secret value remains in Key Vault. The consuming identity still needs permission to retrieve the secret.

Permanent loss of Key Vault objects after accidental deletion. Soft delete alone may still allow purge depending on configuration.

No. Deny blocks matching noncompliant changes. Supported modify or deployIfNotExists policies use remediation tasks and an authorized assignment identity to fix existing resources.

No. Resource locks apply to management operations. They do not stop an authorized blob data-plane deletion; use data permissions and appropriate retention or recovery controls.

In the RBAC model, resource management and secret data access are separate. Grant the appropriate data role, then verify the actual identity, scope, and network path.

Overprivileged standing assignments—prefer eligible PIM roles scoped to what each admin needs.

What an enterprise app or registration can access on behalf of users or the app, and who must approve those permissions.

No. Configure public access separately and validate private DNS from the client. Private connectivity also does not replace data authorization.

Which networks or IPs can reach the storage account's public endpoint, often combined with private endpoints and identity-based access.

Threat protection configurations such as malware scanning and anomalous access detection for storage accounts.

To record database events for investigation and compliance on Azure SQL Database and Managed Instance.

Supported Azure database services with threat detection beyond only enabling TDE.

Applicable subnet and NIC NSGs must allow the flow. ASGs group network interfaces for rules; they are not independent filtering devices.

When you need centralized filtering, FQDN rules, or hub firewall policies beyond simple allow/deny lists on a subnet.

Inspect DNS resolution, private zone links, and forwarding from the affected client. Broadening an RBAC role will not fix an incorrect network destination.

Private connectivity to Azure PaaS or your own services without exposing them on the public internet.

No. Stored access policies apply to service SAS tokens, not user delegation or account SAS tokens. Keep permissions and validity narrow and protect every SAS URL as a credential.

Data overexposure and risks related to Copilot and AI apps, including SharePoint oversharing scenarios.

Identity for AI agents, including Conditional Access and access management for those agent identities.

Microsoft Defender for Cloud Apps, configured through the Defender portal and coordinated with Power Platform. Verify supported agent coverage and protection outputs; a Defender for Cloud AI plan alone is not this integration.

When you need RDP or SSH without a public IP on the VM, through the Bastion service on a dedicated subnet.

No. JIT limits allowed source addresses, ports, and time windows and can protect a private path. It does not create connectivity or authenticate the operating-system session.

To extend Azure security controls to hybrid and multicloud servers, including Defender for Servers scenarios.

No. Agentless machine scanning supports eligible Azure, AWS, and GCP cloud machines. Arc connectivity alone does not give an arbitrary on-premises server that capability; validate its supported protection method.

Stronger VM boot integrity and platform security features on supported security types.

Misconfigurations and runtime risks in container workloads across supported platforms such as AKS.

In front of HTTP applications (for example with Application Gateway or Front Door) to filter common web attacks.

Security posture recommendations and risk identification across connected cloud resources.

Defender for Cloud finds and scores posture and workload threats. Sentinel is the SIEM for correlation, hunting, and orchestration.

To extend posture and protection plans to multicloud environments named in the study guide.

Discovering unprotected internet-facing assets and external attack surface.

Ingests logs and signals from a source into the workspace so analytics and hunting can use them.

The Sentinel service needs permission to invoke the playbook, and the workflow connections need downstream permissions. The analyst also needs the appropriate human role; these permissions are separate.

No. It helps retain available and future data under the configured policy. It cannot recreate deleted events or telemetry that was never collected.

Windows Security events and other telemetry, including paths that use Windows Event Forwarding.

No. Check plugin enablement, configuration, and underlying product permissions. Verify source evidence before acting; enabling a plugin does not guarantee an accurate answer.

No. Configure the required connectors and prerequisites, verify a test event in the destination table, and then enable or tune the relevant analytics content.