Our Linux host patching is under control, but most of our vulnerability backlog is now in container images shipped by development teams. Host patching doesn't fix those images, and ownership is unclear. What tools and processes actually help get the findings fixed?
How do you get container image vulnerabilities out of the ops backlog?
Answers
Start by making ownership explicit. The team that owns the application needs to own updating its dependencies and shipping a rebuilt image. Ops or a platform team can maintain the approved base images and scanning infrastructure, but they can't patch an application image just by updating the host.
Route findings to the team that owns the image, ideally as actionable work rather than a giant shared backlog. You'll probably need management agreement on who handles exceptions and how quickly different findings should be addressed.
Reply to this answer
Put image scanning in the build pipeline. Trivy is one option, and Dependency-Track can complement it. Start with reporting so teams can see what would fail, then agree on thresholds and a timeline before blocking deployments.
You can use different gates for different environments, but make sure there's an exception process. Turning on a hard block against an existing backlog without a plan to fix it will just stop releases.
v.194!cobaltfern:
Known exploited vulnerability lists can help with prioritization too. A CVE count by itself doesn't tell you which finding needs attention first.
Reply to this answer
Make rebuilding boring. Maintain a set of approved base images, refresh them on a schedule, and give teams a straightforward way to rebuild their images from an updated base using the same application source.
Follow the rebuild with automated QA and a deployment process. A fresh base image does nothing for an old image that's still running. Teams needing a different base can request a review or an exception rather than picking one with no shared standard.
Reply to this answer
Don't treat every finding as equally exploitable. Look at the affected package, whether the vulnerable functionality is present and used, and the conditions needed to exploit it. For example, a library may be installed even though the relevant utility or filesystem support isn't used by the container.
That doesn't mean ignoring the finding. Record the reasoning for a lower priority or exception, and revisit it when the image or threat information changes.
Reply to this answer
Scan what's running as well as what gets built. New findings can appear after an image was deployed, so pipeline checks alone won't keep the backlog current. Grype is one option for recurring scans, with alerts routed to the responsible development team.
For a platform evaluation, check whether it can connect findings to images and owners, support pipeline checks, and show useful remediation details. A scanner that only produces another list won't settle the ownership problem.
Reply to this answer
Add your answer
Answers are reviewed before publication.
_e74~parcelwhisk.5:
A ticket is useful, but it needs an owner and enough detail to identify the affected image and package. Otherwise it just moves the backlog to another queue.