On Fri, Nov 29, 2024 at 9:05 AM Marta Rybczynska <[email protected]>
wrote:

> On Fri, Nov 29, 2024 at 3:02 PM Colin <[email protected]> wrote:
>
>> Hi,
>>
>>
>> I would like to use the vector string to do some analysis
>> and prioritization of CVEs and have
>> found that the vector string is better than using the CVSS score. I would
>> argue that in
>> embedded contexts, the environment that a device is used in will result
>> in a different scoring
>> of the vector. The most easy example is a standalone device that does not
>> have any network
>> connections will likely prioritize physical over network attacks, which I
>> believe is opposite to
>> how the CVSS score is determined.
>>
>> I would like for the vector string to be included in the outputs so it is
>> also cached. Post-build
>> tooling could use the NIST API to pull down the vector string, but I
>> would like to avoid hitting
>> the API for what I imagine are the same reasons caching is already done
>> for cve-check.
>>
>> As others have noted in this thread, the NIST API has been rather
>> unreliable lately. I agree
>> that the fetch is likely what is failing the build, since I had to bump
>> the database filename to
>> refresh the contents with the correct vector string. I didn't
>> test_image_json locally, but I did
>> run builds that included fetching the nvd database and everything seemed
>> to have been
>> working.
>>
>>
> Hello,
> I'm not totally convinced by the fact that having the vector will avoid
> you getting the complete entry.
> What is the database if what is needed for the analysis. CVSS scores are a
> little an exception here.
> If we want to change the database format, I think it might make sense to
> add ALL the vector strings
> and not only the last one. That makes way more sense for interpretation
> purposes and allows you
> compare entries that have multiple vectors.
>
> That being said, NVD entries often have multiple vectors - one from the
> vendor issuing the CVE,
> one from the NVD and so on. They often differ quite a bit, and those days
> there might be other
> assignments too. For such an analysis I would rather use the raw CVE
> database instead.
>
> Please note that we keep NVD for now because of the CPEs. Product names in
> the CVE database
> are not always set correctly (also, the vector is not always set). For all
> more complete analysis I
> use CVE that is available as a simple repository at
> https://github.com/CVEProject/cvelistV5
>
> Kind regards,
> Marta
>
>

I don't disagree. I figured including all CVSS vector strings
would probably be best, but I also
figured that would also lead to a breaking API change in the cve-check
output. I was
concerned that a breaking change may not be able to be backported to
Kirkstone and
Scarthgap, which is what I wanted. However, if a patch that adds all CVSS
version vector
strings would be accepted over this patch, I would be happy to make that
change instead.

Thank you,
Colin
-=-=-=-=-=-=-=-=-=-=-=-
Links: You receive all messages sent to this group.
View/Reply Online (#208034): 
https://lists.openembedded.org/g/openembedded-core/message/208034
Mute This Topic: https://lists.openembedded.org/mt/109805499/21656
Group Owner: [email protected]
Unsubscribe: https://lists.openembedded.org/g/openembedded-core/unsub 
[[email protected]]
-=-=-=-=-=-=-=-=-=-=-=-

Reply via email to