Defender for Cloud: A Deeper Dive Across Azure, Multicloud, and On-Premises

Secure score is where most teams start. It shouldn't be where they stop.

The earlier post on Azure Arc and Security covered how Arc projects on-premises and other-cloud servers into Azure Resource Manager so they can be governed like native resources. This post picks up from there and focuses on the security platform that does most of the work once those resources are visible: Microsoft Defender for Cloud.

About 18 months ago, a large share of my work was Defender for Cloud analysis and remediation for an enterprise client. Most of it centered on secure score: reading the recommendations, sorting signal from noise, and working remediation tasks through the Azure portal with resource owners. That experience shapes the practical notes throughout this post.

What Defender for Cloud is

Defender for Cloud is Microsoft's cloud-native application protection platform (CNAPP). It brings together three areas that used to be sold as separate tools:

• Cloud Security Posture Management (CSPM): continuous assessment of configuration and exposure, with recommendations and a secure score.
• Cloud Workload Protection (CWP): threat detection and protection for servers, containers, databases, storage, APIs, and AI workloads, delivered through individual Defender plans.
• DevOps security: visibility into code, pipelines, infrastructure-as-code, and secrets across GitHub, Azure DevOps, and GitLab.

Microsoft positions the platform as covering the full lifecycle from code to runtime. Since May 2026, Defender for Cloud is also generally available inside the Microsoft Defender portal, where posture, recommendations, attack paths, and alerts sit next to the rest of the XDR stack.

Secure score: where most programs begin

The classic secure score in the Azure portal is built from recommendations mapped to the Microsoft Cloud Security Benchmark (MCSB). A higher score means less identified risk. Each recommendation belongs to a security control, and each control is worth a fixed number of points split evenly across the resources it assesses. A resource earns its share only when it passes every recommendation in that control. This is the part that catches teams off guard most often.

Three secure scores, three purposes

Practitioners now run into three different scores, and two of them come from Defender for Cloud. The classic secure score in the Azure portal is the control-based MCSB model described above, covering subscriptions, AWS accounts, GCP projects, and Arc-connected machines. The newer Cloud Secure Score in the Microsoft Defender portal (under Exposure Management) is risk-based: it weighs open recommendations by risk level and factors in asset risk factors such as internet exposure and data sensitivity, plus how critical the asset is to the organization. Microsoft Secure Score, also in the Defender portal, measures posture across Microsoft 365 workloads such as identity, devices, email, and apps.

The two Defender for Cloud scores use completely different calculations, so they won't match even for the same estate. Treating any of these as one number causes confusion when leadership asks why the score didn't move. Pick the one you report on, say which it is, and stick with it.

Foundational CSPM vs. Defender CSPM

Defender for Cloud offers two posture tiers. Foundational CSPM is free and turns on for connected environments. Defender CSPM is a paid plan that adds context and prioritization on top.

Defender CSPM has also absorbed serverless coverage. Serverless protection for Azure Functions, Azure Web Apps, and AWS Lambda became generally available in June 2026 as part of the plan, not as a separate Defender plan. Posture for serverless container workloads (Azure Container Apps, Azure Container Instances, and Amazon ECS on Fargate) followed in July 2026.

Workload protection plans

Posture is half the story. The Defender plans add runtime threat detection and protection, and you enable them per subscription, AWS account, or GCP project:

• Defender for Servers (Plan 1 and Plan 2): Plan 1 centers on Defender for Endpoint EDR and agent-based vulnerability assessment. Plan 2 adds agentless vulnerability, malware, and secrets scanning, file integrity monitoring, just-in-time VM access, and free daily Log Analytics ingestion.
• Defender for Containers: Kubernetes threat detection, container image vulnerability assessment, and Kubernetes data plane hardening.
• Defender for Databases: threat detection and vulnerability assessment for Azure SQL, SQL Server on machines, and open-source relational databases.
• Defender for Storage, Key Vault, Resource Manager, App Service, and APIs: targeted protection for platform services.
• AI workload protection: posture and threat protection for AI models, agents, and applications, an area Microsoft is expanding quickly.

Multicloud: AWS and GCP

Defender for Cloud connects to AWS accounts and GCP projects through native connectors. Once connected, those resources show up in the same inventory, the same recommendations, and the same secure score as Azure. Microsoft lists Defender for Servers, Defender for Containers, Defender CSPM, and Defender for SQL Servers on Machines as generally available for both AWS and GCP. Defender for open-source relational databases on AWS RDS became generally available in June 2026, which also means it's now a billable plan rather than a free preview.

Multicloud posture coverage also expanded on June 30, 2026, when over 200 new AWS and GCP recommendations and about 90 new resource types moved from preview to general availability. These count toward the risk-based Cloud Secure Score in the Defender portal. If that score dropped around then, that's most likely the broader assessment scope, not a worse environment. Call it out before leadership asks.

The practical value is consolidation. Security teams that watch only Azure-native alerts miss whatever runs in another team's AWS account or a data group's GCP project. One posture view and one alert pipeline closes that gap without a second tool or console.

On-premises and hybrid: where Azure Arc comes in

This is where the earlier Arc post connects. Defender for Cloud protects on-premises and other hosted servers primarily by treating them as Azure Arc-enabled servers. (Defender for Endpoint direct onboarding can deliver Plan 1 features without Arc, but you give up the ARM governance below.) Once a machine is Arc-connected, it can:

• Receive Defender for Servers protection, including Defender for Endpoint integration.
• Show up in recommendations and secure score next to cloud resources.
• Be governed with Azure Policy, tags, and RBAC like any other ARM resource.
• Feed SQL Server on-premises into Defender for SQL Servers on Machines.

DevOps security: shifting left

Connectors for GitHub, Azure DevOps, and GitLab let Defender for Cloud surface code-level findings: IaC misconfigurations, exposed secrets, and dependency vulnerabilities. With code-to-runtime mapping (introduced in preview in March 2026, so check its current status), Defender CSPM can trace a runtime finding back to the repository that produced it. The GitHub Advanced Security integration, generally available since May 2026, also pushes runtime context such as internet exposure back into GitHub alerts. That changes remediation from "fix this VM" to "fix the template so this never deploys again."

Running a remediation program that works

Tooling is the easy part. These patterns tend to separate programs that improve posture from programs that only produce reports:

1. Baseline first. Record the starting secure score and recommendation count per subscription before you change anything.
2. Clear resources completely within a control. Partial fixes earn nothing (see the field lesson above), so group work by control and finish each resource you touch.
3. Use governance rules. Assign owners and due dates to recommendations so remediation doesn't depend on one security engineer chasing people.
4. Exempt deliberately. Document risk acceptance or mitigating controls instead of leaving findings open forever.
5. Fix at the source. Use Azure Policy deny/deployIfNotExists and IaC fixes so the same finding doesn't come back next sprint.
6. Prioritize by attack path. Where Defender CSPM is licensed, start with findings that sit on real attack paths to sensitive resources.
7. Report trends, not snapshots. Leadership responds better to a score trend line and closed critical paths than to a raw recommendation count.

Closing thoughts

Secure score is a useful compass, but Defender for Cloud is much broader than one number. Combined with Azure Arc, it gives a single posture and protection layer across Azure, AWS, GCP, and the datacenter. With Defender CSPM, it also tells you which risks deserve attention first. The organizations that get the most out of it treat it as an operating model with owners, governance, and source-level fixes, not a dashboard someone checks once a quarter.

References

• Microsoft Learn: Defender for Cloud multicloud support matrix
• Microsoft Learn: What is Cloud Security Posture Management (CSPM)
• Microsoft Learn: Secure score in Defender for Cloud
• Microsoft Learn: Defender for Servers overview
• Microsoft Learn: What's new in Defender for Cloud
• Microsoft Learn: What is serverless protection?
• Microsoft Learn: Transition from grouped to individual recommendations
• Microsoft Learn: Connect on-premises machines to Defender for Cloud
• Microsoft Security: Microsoft Defender for Cloud
• Microsoft Security: Defender Cloud Security Posture Management

Related reading

• Azure Arc and Security

Views are my own. Feature availability, preview status, and pricing change frequently. Confirm current details on Microsoft Learn before you make design or licensing decisions.