Different checks answer different security questions. Code review can identify an unsafe design, a scanner can detect a known weakness, and artifact verification can reveal that the package being deployed isn't the one you intended to build. Combining them reduces dependence on any single tool or person catching every problem.
Combine prevention, detection, and response across the delivery path. Repeating the same check in several places can improve coverage, but it won't address a weakness they all share.
Put each check where it can change the outcome
Feedback before a commit can prevent a secret or unsafe pattern from entering repository history. Merge request scans help developers examine a proposed change while they still have its context. Default-branch and scheduled checks help find problems that were missed, or vulnerabilities disclosed after the code was written.
A secret found after merge needs a response even if a later change removes it from the file. Revoke or rotate the exposed credential, then address affected copies and access. Deleting the text doesn't invalidate the credential. This is why prevention, later detection, and incident response complement one another.
Choose checks for the application's languages, dependencies, packaging, and exposure. Static analysis, dependency scanning, container scanning, and dynamic tests don't cover identical risks. A dynamic check also needs a suitable running target; simply enabling another scanner doesn't establish that it tested the application you deliver.
Review scan results before expanding coverage
Try scanner configuration with projects that represent different technologies, repository sizes, release paths, and risk. Check that scans finish and give developers enough information to investigate the findings. Include runner capacity and access to analyzers or advisory data, particularly in restricted networks.
This pilot also gives the people handling findings a realistic view of the workload. Decide who reviews findings, who fixes them, and who can accept an exception as results arrive. A large rollout without a follow-up process creates a backlog faster than it reduces risk.
Move proven shared configuration into a versioned CI/CD component or library when it helps teams adopt consistent settings. Required coverage needs enforcement as well: a project can remove a normal include or change its local pipeline. Scan execution policies enforce supported GitLab scans; pipeline execution policies support custom jobs and third-party scanning workflows.1 Verify the effective jobs and reports in the projects covered by the policy.
Every layer consumes time and operating capacity. Weigh the problems scans catch against false positives and pipeline delays. A slower check can still be worthwhile for an exposed service, while an irrelevant scanner can consume resources without improving coverage.
Security scanning that covers what you ship explains how to choose a starting set of applications, examine the first results, and expand coverage with the people who will maintain it.
Protect the software and the records you rely on
Review the inputs to a build as carefully as its application code: shared pipeline configuration, build images, dependencies, and package sources all influence the result. Use trusted sources and versioned inputs so teams can identify a change and investigate it later.
Keep a software bill of materials, or SBOM, associated with the software it describes. Check whether it includes indirect dependencies and how supplier components are represented. That connection helps locate affected releases when a new advisory arrives. An SBOM records components; it doesn't establish that they're safe.
Artifact digests identify content. Provenance describes where an artifact came from and how it was built. Verify that information against the expected source and trusted build process when accepting an artifact.
Protect the generation and handling of scanner reports too. Merge request approval policies evaluate report contents; they don't verify a report's integrity or authenticity. Restrict who can alter pipeline and policy configuration, and consider what a compromised execution environment could change.
Teams also need somewhere to track the resulting work. GitLab's vulnerability report supports that workflow for default-branch findings,2 while teams using scanner artifacts directly need another way to retain owners, decisions, and fixes. Basic SAST and downloadable reports are available without integrated vulnerability management.3
Further Reading
Notes
-
Scan execution policies and pipeline execution policies are available in Ultimate. They enforce centrally managed checks across projects; reusable configuration alone doesn't prevent a project from removing a check. ↩
-
The vulnerability report is available in Ultimate. It provides a shared view for triage and tracking; downloading scanner artifacts doesn't provide the same workflow. ↩
-
Basic SAST scanning and report artifacts are available on all tiers. GitLab Advanced SAST and integrated vulnerability management are available in Ultimate. ↩
- Choose application security checks explains the available testing approaches and their coverage.
- Prevent, detect, and respond to exposed secrets helps place secret checks through the workflow.
- Compare scan execution policies with pipeline execution policies when applying required checks across projects.
- Secure the pipeline and its inputs and verify build provenance explain how to protect the software supply chain.
- Review dependencies and export an SBOM describes GitLab's dependency list, available with Ultimate.