Vulnlog and your scanner
Your scanner finds vulnerabilities in your dependencies, and Vulnlog records what you concluded about them. The two meet in the suppression file: Vulnlog writes it, and your scanner reads it.
What Vulnlog and the scanner each do
Vulnlog doesn’t scan anything. Your scanner keeps finding vulnerabilities, with its own database and its own rules, and you keep running it as you do today. When it reports something new, you analyse the finding and record the result in the Vulnlog file, with a report that names the scanner.
The scanner, in turn, knows nothing about your analysis. It only learns which findings to stay quiet about, from the suppression file Vulnlog generates for it. For scanners with a native format, such as Trivy, that’s a file in the scanner’s own format, under the file name the scanner looks for. For the other scanners, Vulnlog writes a generic JSON file that you can feed into your own tooling. See Use generic suppression files. Supported scanners lists the formats and file names.
From one entry to the scanner
Text4Shell in Acme Shop came from OWASP Dependency-Check.
The team found that Acme Shop never calls the vulnerable code and recorded the verdict not affected.
From that entry, Vulnlog generates dependency-check-suppressions.xml, which lists CVE-2022-42889 with the analysis as the note.
For the generated file, see Suppression.
The next Dependency-Check run doesn’t report Text4Shell again. Anyone who wonders why can read the note, or the entry in the Vulnlog file.
When several scanners report the same vulnerability, the entry has one report for each of them, and each scanner’s suppression file lists only the identifier types that the scanner accepts.
If a scanner calls the vulnerability by its own identifier, you give that identifier on the scanner’s report, and the scanner’s file uses it.
Snyk only accepts SNYK- IDs and cargo-audit only RUSTSEC- IDs.
If the entry’s id is a CVE and the report’s vuln_ids has no such ID, the entry is left out of that scanner’s file.
For the identifier types of each scanner, see Supported scanners.
When you add or replace a scanner, you add reports for it to the entries it flags, and Vulnlog generates its file.
The analysis stays as it is.
Which findings stay quiet
Vulnlog decides from your entries which findings go into a suppression file.
An entry with the verdict not affected goes in, because the vulnerability can’t harm your software and the scanner would only repeat itself.
That’s what happens with Text4Shell.
An entry with the verdict affected, or one still under investigation, goes in only if you ask for it with a suppress block on the scanner’s report.
A vulnerability you haven’t fixed is a real risk, so it stays visible until you decide otherwise.
Spring4Shell in Acme Shop is affected, and the fix is planned for 4.4.0.
Its Trivy report has a suppress block that expires on 2027-03-31, so Trivy stays quiet until then and starts reporting it again if the fix hasn’t shipped.
The lodash entry is an accepted risk, and its npm audit report has a suppress block without an expiry date.
Because npm audit has no suppression format of its own, Vulnlog writes the entry to npm-audit.generic.json, which npm audit doesn’t read.
Your own tooling applies it.
A resolved entry never goes in, whatever the verdict.
After Log4Shell was fixed in 4.3.0, no suppression file contains it, so the scanner confirms that the fix works.
This applies as soon as you record the resolution, even if its release hasn’t shipped yet.
A scanner that checks the current release then reports the finding again, because that release still has the vulnerable package.
With --as-of and the release the scanner checks, the resolution doesn’t count yet, and a suppress block on the report still applies.
The exact rules, including expiry, are in Suppression.
Generated files follow the Vulnlog file
Vulnlog writes a scanner’s file even when it has nothing to suppress, so an outdated suppression can’t stay behind after you resolve a vulnerability or a suppression expires.
In Acme Shop, grype.generic.json is empty, because Grype only reported Log4Shell, which is resolved.
If you remove the last report of a scanner from the file, Vulnlog stops writing that scanner’s file, and you delete the old one yourself.
Reports from other, such as a customer inquiry, don’t name a scanner, so they don’t get a suppression file.
Because suppression files are generated, an assessment changes in the Vulnlog file and reaches the scanner the next time Vulnlog writes the files. When the files are generated in the same pipeline that runs the scanners, each scan uses the assessments in the current version of the Vulnlog file.