nscuro / gotalias

Got alias?

Got Alias?

Usage

docker compose up -d # Launch Neo4j

go run main.go \
  -github-token <GITHUB_PAT> \
  -osv \
  -ossindex-user <USERNAME> -ossindex-token <TOKEN> \
  -snyk-orgid <ORG_ID> -snyk-tokens <TOKEN1>,<TOKEN2>,<TOKENX> \
  -purls purls.txt

Note
Data from vulnerability databases that can not be mirrored has to be requested on a per-component basis. A list of PURLs can be provided for this purpose. To generate a PURL list from multiple CycloneDX BOMs in JSON format:

for f in ./**/*.json; do jq -r '.components[].purl' $f >> purls.txt; done
sort -u purls.txt > purls_unique.txt

Useful Cypher Queries

// Find a specific vulnerability by its ID
MATCH (:Vulnerability {id: "GHSA-5mcr-gq6c-3hq2"})
RETURN *

// Find the top 100 most aliased non-Debian vulnerabilities
MATCH (vuln:Vulnerability)<-[r:ALIASES]-(:Vulnerability)
WHERE NOT vuln.id STARTS WITH "DLA-"
WITH vuln, COUNT(r) as aliasCount
RETURN *
ORDER BY aliasCount DESC
LIMIT 100

// Find all Snyk vulnerabilities and their aliases
MATCH (v:Vulnerability)-[:ALIASES]->(:Vulnerability)
WHERE v.id = STARTS WITH "SNYK-"
RETURN *

// Find all Snyk vulnerabilities that alias more than one CVE
MATCH (v:Vulnerability)-[r:ALIASES]->(a:Vulnerability)
WHERE v.id STARTS WITH "SNYK-"
    AND a.id STARTS WITH "CVE-"
WITH v, COUNT(r) AS rc
WHERE rc > 1
RETURN *
ORDER BY rc DESC

// Find all vulnerabilities for which more than one alias was reported by GitHub
MATCH (v:Vulnerability)-[r:ALIASES]->(:Vulnerability)
WHERE ANY(item IN r.reportedBy WHERE item = "GITHUB")
WITH v, COUNT(r) AS rc
WHERE rc > 1
RETURN *
ORDER BY rc
LIMIT 100

Data Samples

GHSA-qcvw-h34v-c7r9

Multiple CVEs have been registered for the same GHSA ID.

The official GitHub advisory only refers to CVE-2021-43299, but 4 more CVEs link back to the GHSA ID:

• CVE-2021-43300: Stack overflow in PJSUA API when calling pjsua_recorder_create
• CVE-2021-43301: Stack overflow in PJSUA API when calling pjsua_playlist_create
• CVE-2021-43302: Read out-of-bounds in PJSUA API when calling pjsua_recorder_create
• CVE-2021-43303: Buffer overflow in PJSUA API when calling pjsua_call_dump

OSV thus reports all those CVEs as aliases for GHSA-qcvw-h34v-c7r9.

Because GHSA-qcvw-h34v-c7r9 aliases CVE-2021-43299, does it mean that it aliases the other 4 CVEs, too? And does it imply that CVE-2021-43299 and CVE-2021-43303 are also identical? Technically, the other CVEs describe different things than CVE-2021-43299.

GO-2022-0586

The aliases field¹ in OSV is misused² by some vulnerability databases. Instead of populating this field with identifiers of vulnerabilities that are truly identical, it is populated with related³ vulnerabilities.

GO-2022-0586 is an advisory rather than a vulnerability, as it represents an aggregate of multiple vulnerabilities.

While there are four pairs of CVEs and GHSAs that indeed alias each other, none of the CVEs or GHSAs actually alias GO-2022-0586; They are simply not the same. It also can not be assumed that any of the GHSAs alias each other. The same is true for the CVEs.

CVE-2021-21290

At the time of writing, Snyk has a total of 12 SNYK- vulnerabilities assigned to CVE-2021-21290 / GHSA-5mcr-gq6c-3hq2. It appears like Snyk created separate vulnerabilities for each affected netty module. The vulnerabilities differ only in their How to fix? advice.

Note
Snyk explicitly does not state that any of the vulnerabilities it reports in the problems field of its PURL API⁴ are aliases. It just so happens that oftentimes they are. But there's no guarantee.

Footnotes

1. https://ossf.github.io/osv-schema/#aliases-field ↩

2. https://github.com/google/osv.dev/issues/888 ↩

3. https://ossf.github.io/osv-schema/#related-field ↩

4. https://apidocs.snyk.io/?version=2023-03-29#get-/orgs/-org_id-/packages/-purl-/issues ↩