Microsoft SC-500: Defender for Cloud Attack Paths
Defender for Cloud attack paths help prioritize combinations of exposure, permissions and vulnerabilities that lead toward critical assets. The engineering work begins with validating the graph and ends only when a real exploit chain is broken. A credible assessment follows a possible attacker from an entry point to a sensitive resource, checking each identity and network relationship along the way. Remediation should remove or contain that route and then demonstrate that the same path no longer works.
On this page
- Identify which assets attackers can reach
- Understand cloud security graph gaps
- Trace exposure through identity, network and data layers
- Prioritize by impact, exploitability and remediation control
- Investigate paths with Cloud Security Explorer
- Coordinate different owners along one attack path
- Turn repeated exposure into platform guardrails
- Record uncertainty as part of graph-based prioritization
- Prove attack paths are broken
Ask which protected asset an attacker could actually reach
A long list of posture findings hides which problems combine into serious risk. Microsoft Defender for Cloud attack-path analysis uses graph context about resources, identities, vulnerabilities, network reachability and sensitive targets to identify sequences that could be exploited. A public VM with a modest vulnerability may matter far more if its attached identity can reach sensitive storage than a more severe vulnerability on an isolated test system. Risk is shaped by the reachable consequence as well as the technical weakness, so remediation should begin with the business asset that requires protection.
A useful analysis records the external entry point, every intermediate identity or network relationship and the target asset’s owner. Confirm the graph’s evidence before declaring the path exploitable in the organization’s actual environment. A relationship may depend on a permission no longer in use, a firewall rule created for testing or data that is less sensitive than its label suggests. Validation makes a posture graph an engineering decision aid rather than a source of unquestioned severity scores.
Understand the cloud security graph and its blind spots
The cloud security graph represents relationships among cloud resources, permissions, exposure and other supported security signals. Defender CSPM and relevant onboarding or scanning capabilities determine which information appears. An absent path does not prove there is no attack route: a disconnected subscription, a missing agentless scan, a service outside supported graph coverage or stale inventory can hide dependencies. Attack-path analytics is therefore only as complete as the resource and identity telemetry it receives.
Before prioritizing based on the graph, compare its resource count with the platform inventory and inspect one representative high-value workload. Does the graph show the attached managed identity, network boundary, sensitive database and relevant vulnerability? If not, record that coverage gap and determine which onboarding or collection prerequisite is missing. Avoid presenting ‘zero detected attack paths’ as a security success without explaining the confidence and supported scope of the detection.
Trace exposure through identity, network and data layers
A common path begins with reachable code execution or stolen credentials, continues through an overprivileged service identity and ends at a storage account or database. The exposed host, Azure RBAC assignment and data authorization are distinct edges. Removing one path does not always require rebuilding the whole workload: a narrowly scoped role may stop access to the high-value target while a patch is tested, or a correct private endpoint and firewall design may remove unnecessary external reachability. The engineer must choose the change that addresses the meaningful risk.
For a safe lab, grant a test application’s identity excessive read permission to a nonproduction data store, then model how compromise of its host could misuse that authority. Reduce the permission to only the application-specific data action and re-evaluate the path. Verify the effective deny using the principal itself. That test is stronger than relying solely on the disappearance of a visualization, because the authorization change is independently demonstrated at the target service.
Prioritize by impact, exploitability and remediation control
Not every discovered path deserves the same urgency. The target’s sensitivity, external exposure, exploit maturity, privilege breadth and operational consequence shape priority. A graph edge involving a disposable dataset may have a lower consequence than one that reaches backups or a production control plane. Some remediation choices are urgent and reversible, such as narrowing a public allow rule, while others need planned maintenance or code change. Risk ownership should include the business owner who can judge consequence, not only the security analyst who reads the graph.
A triage record should explain what remains exposed if the team fixes only one weakness. If patching an exposed server takes several days, temporary access restrictions may break the path, but those restrictions need ownership and expiry. If an attached identity remains broad after the network change, the risk can reappear through another entry point. Choose improvements that reduce the likelihood of similar paths being recreated, rather than optimizing a single ticket’s closure time.
Use Cloud Security Explorer for questions beyond prebuilt paths
Cloud Security Explorer can query supported graph relationships to investigate exposure hypotheses that are not necessarily represented as a highlighted attack path. Security teams can look for identities bridging subscriptions, services with public exposure near sensitive data or resources whose weakness is repeated across several landing zones. A focused query should correspond to a clear question; browsing every relationship until something looks alarming often produces noise and confuses coincidence with exploitable privilege.
Create a saved investigation for an explicit class, such as internet-reachable compute with an attached identity capable of accessing a critical datastore. Verify a sample of results by checking the real role assignment and network configuration. Share the query assumptions with platform owners so the result becomes reproducible posture evidence. When the organization’s security model changes, review the query to ensure it still reflects real asset classes and supported telemetry instead of silently returning incomplete results.
Coordinate different owners along one attack path
An attack path can cross network, IAM, server, application and data teams. Assigning the whole chain to one generic security queue often delays resolution because no single owner controls every edge. Break the remediation into precise actions: the network team can close an unintended public route; an identity owner can narrow the role; the workload team can patch or redesign the vulnerable service; and the data owner can confirm sensitivity and acceptable access. One coordinating owner must still track the end-to-end result.
At a review meeting, map each proposed fix to its owner, change window, proof method and rollback. Where two controls can break the same path, compare operational cost and durability. A firewall rule that depends on an opaque exception may be less maintainable than correcting an application identity, whereas a temporary network restriction may be the quickest containment during active exploitation. The important metric is the remaining route to the protected asset after every planned change.
Turn repeated exposure into platform guardrails
Attack-path findings are especially valuable when multiple workloads repeat the same dangerous deployment pattern. If each new project creates a broad managed-identity role or leaves a public storage endpoint available, remediating individual resources will not scale. Azure Policy, approved templates and least-privilege role patterns can prevent the underlying resource state from appearing again. Defender for Cloud becomes part of an engineering feedback loop when a finding leads to a durable change in how workloads are provisioned.
Measure recurrence after the template or policy fix. Deploy one representative compliant workload and one intentionally noncompliant candidate in a safe environment, then verify the intended allow and deny behavior. Confirm that legitimate applications continue to work and that the new guardrail does not hide an exception in another subscription. Record any exemption with expiry. Closing many historic paths is good; preventing the same path on every new deployment is better.
Record uncertainty as part of graph-based prioritization
Security graph data can be incomplete or delayed even when the Defender portal shows a plausible path. Record which edges were directly confirmed, which depend on assumptions and which require additional evidence from resource owners. A suspected reachable endpoint might sit behind an undocumented proxy; a role might be technically granted but unavailable because the identity has been disabled. Confidence should therefore travel with the remediation recommendation, not disappear when the finding is copied into a priority queue.
Define a short validation plan with an assigned owner and time limit for uncertain high-risk paths. Use safe read-only checks of effective rules, role assignments and resource identity first, then controlled negative tests where authorized. Mark a path unconfirmed if required visibility is missing; do not downgrade it to harmless simply because the graph cannot complete every edge. This makes prioritization defensible and helps teams improve the coverage that the next assessment relies on.
Prove that the path was broken and the response can use the evidence
The final step is verification from more than one source. The Defender graph should reflect changed exposure or permission, while a direct access test should confirm that the attacker’s hypothetical principal no longer reaches the target. Monitoring should also expose attempted misuse so analysts know whether a blocked operation is occurring. When an incident is active, the graph can help estimate possible lateral movement, but investigators still need timestamps and direct event evidence before concluding that a path was used.
Capture a before-and-after record with initial chain, chosen edge, changed control, negative test, source evidence and remaining risk. Feed the result into incident runbooks and architecture review so responders can quickly identify the same privilege or network relationship in a future case. An attack-path program should report material business risks reduced and recurring patterns fixed, not only the number of colorful lines removed from the dashboard.
After a suggested attack-path remediation, capture the before and after evidence at the individual control layer rather than relying only on the disappearance of a dashboard tile. If the mitigation narrows a workload identity, run an authorized access test from that identity to the critical destination and verify the result is denied. If it restricts a public endpoint, verify the effective exposure from an appropriate external test location and confirm legitimate private callers still work. Record the time the change was made, the responsible owner, the resource identifiers and the next posture refresh time; graph updates may not be instantaneous.
A missing path can have multiple explanations. A scanning connector might have stopped reporting, an asset could have been decommissioned, or the underlying relationship might truly have been removed. Compare source coverage and control evidence before declaring risk eliminated. This distinction is particularly important for operational metrics: security should not look better merely because the graph has lost the data needed to see the problem.