What Vulnlog leaves out

The Vulnlog file has no field for a CVSS score, a CWE category, an advisory text, or the affected version ranges of a vulnerability. The file records what only your team knows, and leaves out what anyone can look up.

Public data has a home already

Everything an advisory database tells you about a vulnerability is public, and you find it by the vulnerability’s identifier. Scores get revised, advisories get updated, and new affected versions are added. A copy in your Vulnlog file would start to go out of date the day you made it, and nobody would notice.

So an entry keeps the identifier, any aliases, and a short name and description that help a reader recognise the vulnerability. The id of Log4Shell, CVE-2021-44228, and its alias, GHSA-jfh8-c2jp-5v3q, are enough to look up its advisories. Most scanners show the scores and advisory details as well.

What no database knows

No database can tell you whether a vulnerability matters for your software. Only your team can, after reading the code. That’s what the file is for: your analysis, your verdict, your plan, and the release that fixes the vulnerability.

The Text4Shell entry in Acme Shop shows the difference. The advisory describes a remote code execution in Apache Commons Text. The entry says something the advisory can’t: Acme Shop never calls the vulnerable code, so it’s not affected.

Severity is your assessment

An entry with the verdict affected has a severity. It’s your rating of how much the vulnerability harms your software, in the way you build and run it, and it can differ from the CVSS score. Acme Shop rates Spring4Shell medium, because the published exploit needs a WAR deployment on Apache Tomcat, and Acme Shop runs as an executable JAR. The advisory can’t take that into account. When a public score influenced the rating, the analysis can say so.

Findings without a public ID

An entry needs an id that starts with CVE-, GHSA-, RUSTSEC-, or SNYK-. These are the IDs that scanners report and that your customers' tools can match. A flaw in your own code that a penetration test, an audit, or a security researcher found has no such ID, so it can’t be an entry. Track it in your issue tracker. If the flaw later gets a public ID, for example because you publish a security advisory for it, you can record it from then on.

A customer or a researcher can still be the source of an entry. When they report a vulnerability that has a public ID, the entry gets a report with the reporter other.

Workflow and history stay in other tools

The file also has no assignee, due date, discussion thread, or change history. Who worked on an entry, and when, is in the Git history. Planning the work belongs in your issue tracker.

Because an entry holds your reasoning and leaves the advisory where it is, it stays short enough to review in a pull request. It needs updating only when your own conclusions change, and an updated advisory leaves it as it is.