Treat the advisory as the start of an investigation
A vulnerable dependency advisory deserves attention, but it does not yet contain the whole decision. Begin by identifying the package, installed version, issue identifier and the release containing the fix. Record which application and deployment use the dependency as well. Without that link, a team can update the wrong repository or close an alert while an older version remains in production.
OWASP recommends analysing dependencies from the beginning of a project and maintaining visibility of issues as they emerge. Monitoring does not replace analysis of the specific case. A high score can help with priority, but it does not say on its own whether the affected function is present, reachable from the internet or dependent on a configuration the application does not use. The first aim is to turn a generic alert into a testable question about a real website release.
Keep a minimum record before changing anything
Capture the evidence before updating: component name and version, the direct dependency that brings it in, the supplier advisory, affected releases, available fix and detection date. Add the application version or deployment identifier. Do not copy secrets, environment variables or user data into this record; identifying the system and the component's technical path is enough.
A release-specific inventory makes this work easier. OWASP's guidance on SBOMs and dependency graphs distinguishes direct from transitive components and recommends linking an inventory to every release. A dependency manifest describes the project's intent, but the final artefact may also contain bundled libraries or operating-system packages. It is therefore useful to know what was built and what actually reached the environment, as well as what the repository declares.
- Advisory identifier and primary source.
- Component, version and direct or transitive relationship.
- Application, published release and affected environments.
- Status: pending, temporarily mitigated, fixed, or not affected with justification.
Check whether the application reaches the affected function
Find the direct dependency that introduces the component and where it is used. Then compare the advisory with the application's behaviour: vulnerable function, input type, required configuration and available permissions. An installed package does not prove that the flaw is exploitable; the absence of an obvious call does not prove that it is isolated either. It may be involved through a plug-in, background job or transitive route.
Hypothetical example: a file-processing library receives an advisory concerning one particular format. The public website displays only images, but an administrative task reuses the library to import documents. The public route may not reach the function, while the import still requires investigation. The useful conclusion is not to label the whole website safe or vulnerable. It is to record which journey was checked, which inputs it accepts and which part remains unresolved.
Choose a response you can verify
When the supplier has released a fixed version, update it in a test environment first. Review the release notes and incompatible changes; a major upgrade may require altered calls, configuration or server requirements. OWASP proposes running the unit, integration, functional and security tests connected to the dependency before taking the correction into production.
If no fixed version exists yet, a temporary mitigation is valid only when it blocks the issue's specific conditions. It might disable a feature, restrict an input or isolate a process, but it must be tested and linked to a permanent correction. Ignoring the advisory without a defined scope, or changing the declared version so that a tool stops matching it, does not fix the component. When a project cannot adopt the new release, it should assess a maintained patch, replace the dependency or plan the necessary adaptation according to the risk and available technical capability.
Test the correction and prepare a way back
Testing must answer two different questions: whether the vulnerable condition has stopped working and whether the website still performs its intended job. Where possible, create a regression test that fails with the vulnerable version and passes with the corrected one. Add the journeys that rely on the component, such as document generation, file uploads, authentication or communication with external services. A scanner no longer displaying the alert is useful evidence, not a complete behavioural check.
Before deployment, decide how to confirm the effective version and what to do if an incompatibility appears. Preserve the previous artefact and a recoverable copy of data when the change could alter its format. Rolling back may restore service, but it can also reintroduce the vulnerability. It should therefore be a limited measure while the compatibility fault is corrected or a tested mitigation is applied.
Document exceptions without turning them into forgotten alerts
A tool may continue to flag a component when a fix has been backported to an older release, or when evidence shows that a particular condition does not affect the product. If an exception is accepted, restrict it to the exact advisory and record who reviews it, which release supports the conclusion, the evidence behind it and when it must be reassessed. Do not exclude the entire package, because a different advisory may appear later.
After deployment, confirm the running version, repeat the relevant test and update the release inventory. Advisories evolve as new information appears, and a dependency that is unreachable today may become reachable after an application change. Making this review part of website maintenance preserves the decision history and supports a faster response without promising absolute security.
- Does the alert match a version that is actually deployed?
- Have the direct or transitive route and its use been identified?
- Was the correction or mitigation tested against the affected condition?
- Does the website retain its essential journeys after the change?
- Does any exception have a defined scope, evidence and review date?
Sources and documentation
- OWASP: Vulnerable Dependency Management Cheat Sheet ↗
- OWASP: Dependency Graph & SBOM Best Practices Cheat Sheet ↗
Content created with AI assistance. Check the sources, date and scope of each guide.