Build stage
Base images, pinned versions, package provenance, multi-stage builds, build secrets, and unnecessary tooling.
Analyse container risk from the image build stage through registry, deployment, orchestration, runtime, secrets, networking, and monitoring.
The strongest coursework connects build-time weaknesses with deployment and runtime consequences. A vulnerable image, excessive privilege, exposed secret, or permissive Kubernetes role matters because of how it changes the attack surface of the running workload.
Container security is easier to explain when the report follows the lifecycle. Start with the Dockerfile and dependencies, move to the image registry, then examine deployment configuration, orchestration controls, runtime behaviour, and monitoring. This avoids treating every scanner alert as an equally important issue.
For each stage, identify the asset and trust assumption. A build pipeline may trust a base image; a registry may trust an identity; a Kubernetes workload may trust a service account. Security analysis becomes stronger when those assumptions are visible.
Base images, pinned versions, package provenance, multi-stage builds, build secrets, and unnecessary tooling.
Resource limits, read-only filesystems, dropped capabilities, non-root users, secret injection, and policy enforcement.
Network exposure, lateral movement controls, audit logs, anomalous process behaviour, and incident containment.
Kubernetes assignments often include a large YAML file and ask for a security review. Instead of commenting on syntax line by line, group observations by identity, privilege, network isolation, secret handling, workload configuration, and observability.
For example, an overly broad ClusterRole should be discussed in terms of the permissions it grants and the blast radius if the associated identity is compromised. A missing NetworkPolicy should be tied to unnecessary east-west connectivity rather than described only as a missing object.
If the brief requires architecture discussion, compare namespace boundaries, workload identities, ingress controls, and the responsibilities of the cluster versus the application.
Show which principal receives which permissions and whether least privilege is maintained.
Review privilege, capabilities, host access, filesystem settings, seccomp, and image provenance at a conceptual level.
Explain how allowed communication should follow application dependencies rather than default-open connectivity.
A vulnerability scanner can produce hundreds of results. Your report should separate raw severity from practical risk. Consider whether the affected package is actually present in the final image, whether the vulnerable feature is reachable, whether compensating controls exist, and whether a fixed version is available.
A concise table can record component, finding, evidence, likelihood, impact, remediation, and validation. This is more useful than copying a scanner screen into the appendix without interpretation.
For pipeline controls and automation, see the dedicated DevSecOps page. For cloud-hosted clusters, the cloud security page can help connect workload controls with IAM and platform logging.
Capture only the evidence required by the brief: relevant Dockerfile lines, selected YAML fragments, scanner summaries, policy outputs, and before/after configuration where a fix is demonstrated. Redact secrets and avoid exposing credentials in screenshots.
Finish by explaining how the recommended control would be verified. A remediation is stronger when the report includes a measurable validation step such as confirming the workload runs as non-root, a denied RBAC action, a blocked network path, or a clean rebuild with the updated dependency.
The report identifies image source, dependencies, and important build assumptions.
Identity, capabilities, host access, and service-account permissions are tied to real workload needs.
The final section explains how each important change can be checked after implementation.
Yes. Topics can include RBAC, service accounts, NetworkPolicy, workload identity, namespaces, admission controls, pod security settings, secrets, and audit evidence.
Yes. Guidance can explain package findings, base-image choices, severity, exploitability context, remediation, and the difference between an image finding and an observed runtime exposure.
Yes. Image scanning, dependency checks, IaC review, policy gates, signed artifacts, and deployment verification fit naturally into a secure CI/CD discussion.
Yes. Practical work should use local containers, course clusters, sandboxes, or other systems you are authorized to test.