A vulnerable dependency was found: what to do next
Triage a vulnerable dependency, assess its use and plan updates, testing and follow-up rather than assuming a passing build resolves the risk.
Most projects use third-party libraries for things like document generation, HTTP and parsing. That saves time, but your app's security now partly rests on code you did not write. OWASP's Vulnerable Dependency Management Cheat Sheet is a decision guide for what to do when a scanner flags a dependency. It says its proposals are not silver bullets and should be adapted to your context. We read all of it on 15 November 2026 and not the OpenSSF, Debian or Red Hat pages it links. Everything attributed is OWASP's. Tool names appear only because the sheet lists them.
Start early
The sheet recommends automated dependency analysis from the start of a project. Adding it mid-project or late can create a large backlog that burdens the team and may block progress.
How flaws become public
- Responsible disclosure: a researcher works with the provider and a CVE identifier is issued with the available fix or mitigation. Disclosure timelines vary; follow the applicable coordinated-disclosure policy rather than assuming a universal 30-day rule.
- Full disclosure: details and exploit code are published, and a CVE may not exist, so scanners that rely on CVE data can be blind to it. Pick a scanner that uses several sources and lets you mark false positives.
The five cases
The sheet says that when the flaw is in a transitive dependency (a dependency of your dependency), act on the direct dependency, because acting on the transitive one often affects stability and requires a full understanding of the chain.
- A patched version exists. Update in a test environment and run your tests. If they pass, ship. If code needs changing for the new API, change it and retest. If the fix cannot be adopted (breaking major version, a pinned transitive version, or a newer runtime requirement), raise it with the provider and use the interim steps in case 2, or case 5.
- The provider says a fix will take months. Apply any workaround they give. Otherwise find the calls that reach the affected functions and the conditions needed to exploit them. The sheet allows a wrapper that checks input only if it blocks those conditions on every affected path; otherwise disable or isolate the feature. Treat it as temporary. Use any provider-supplied exploit as a regression test, but do not read a blocked payload as proof you are safe. Note the CVE in your README, because the scanner will keep alerting, and scope any ignore entry to that single CVE.
- The provider will not fix it (or does not answer). This is the sheet's last resort. If the library is open source, you can patch and contribute the fix. Start in parallel looking for a better-maintained component, or if it is a commercial product, press the supplier. Use the CVE text and the upstream advisory to learn what actually needs to change, and write a test that reproduces the flaw. OWASP's bar for an in-house patch: the test fails on the unpatched version and passes on the patched one; the dependency's own tests still pass; and only then suppress the alert, per CVE. Changing a version string so the scanner stops matching "records a fix but does not make one".
- You found it, not the provider (a full-disclosure post or a penetration test). Tell the provider; if they cooperate, continue as in case 2, otherwise as in case 3, using the post or test report instead of the CVE.
- A fix exists but you cannot adopt it (major version breaks your code, a transitive pin, or an unmaintained version line). The sheet describes backporting the fix to the version you run. It says to confirm the upgrade is really blocked by trying it, record the blocker, keep the patch minimal, verify it as in case 3, publish the patched build through the registry your builds already trust with traceable provenance, and treat it as an ongoing commitment, since each new upstream release and CVE means more patching. Drop the backport as soon as the fixed version is usable.
Who decides to accept a risk
The sheet says accepting the risk of a known issue should be a decision of the company's Chief Risk Officer (or the security chief), based on technical analysis and CVSS scores. In a small company, the useful point is that someone with authority makes this decision explicitly and records it, rather than a developer silencing an alert.
Choosing detection and remediation tools
The sheet lists free and commercial scanners and a tool for finding old dependencies, and says tools find problems but do not fix them. For remediation, it points to Linux distribution security teams, which backport fixes into stable releases, as the model, and for language packages it says the choice is between maintaining patches in-house and paying a provider that does. It suggests judging a paid option on coverage, published response times, published patches with provenance, delivery through your existing registry, and a documented way out. The sheet is written by OWASP community contributors and we have not evaluated any provider or tool.
Questions for your team
- Do we scan dependencies automatically on every build, and who sees the results?
- When an alert fires, who decides: upgrade, workaround, replace or accept?
- Do we have tests covering the features that use each dependency?
- Is any suppression broader than a single CVE?
- Is any dependency unmaintained?
Related: Docker container security checklist, secrets management basics, cloud cost control.
Limits
One OWASP page, read in full; tool support lists and the remediation section reflect the sheet on the fetch date and may be out of date. We give no code, statistics or tool recommendations. Decisions about accepting risk in a regulated business may involve legal and compliance advice beyond this post.
By Omnistu.