Missing advisory for CVE-2023-52251 (provectus/kafka-ui) — 447M-pull image with an unpatched RCE and no matchable entry anywhere

Bắt đầu bằng cách xem xét các mục advisory hiện có, gồm cả mục đã được xem xét và chưa được xem xét, cùng với GitHub Advisory Database API, sau đó so sánh các trường hệ sinh thái được hỗ trợ và gói bị ảnh hưởng của chúng với thông tin chi tiết về Docker image trong issue này. Kiểm tra SECURITY.md và bằng chứng được tham chiếu trong MessageFilters.java, bao gồm khoảng >= 0.4.0 và <= 0.7.2. Được xem là hoàn tất khi repository đã quyết định liệu artifact không được hỗ trợ này có thể được biểu diễn hay không và, nếu có, đã chấp nhận dữ liệu advisory theo schema dự kiến.

CVE-2023-52251 (CVSS 3.1 8.8 HIGH, published 2024-01-25) has no entry in the GitHub
Advisory Database. I checked all three types via the API — reviewed, unreviewed and
malware all return zero hits for this CVE ID.

It is a remote code execution issue in provectus/kafka-ui: the "smart filter" evaluates
attacker-supplied Groovy via the q parameter of
/api/clusters/local/topics/{topic}/messages. A public exploit exists.

Why I think this one is worth carrying even though it is not a package:

• The affected artifact is the Docker image provectuslabs/kafka-ui, with
  447,220,745 pulls and Docker Hub status active.
• The repository has 12,278 stars and 1,411 forks and is not archived, with no commit
  since 2024-04-08.
• Its own SECURITY.md still claims 0.7.x is "currently being supported with security
  updates".
• No other database produces a matchable entry either: the MITRE record has
  vendor/product/version = n/a, NVD derives only
  cpe:2.3:a:provectus:ui:*:*:*:*:*:kafka:*:* with no version boundaries, and the OSV
  entry has an empty package field. So Dependabot, Docker Scout, Trivy and Grype are all
  silent on it.
• It is exploited in the wild. An operator running provectuslabs/kafka-ui:latest exposed
  without authentication had the container used to mine cryptocurrency for about 1.5 days,
  resulting in a hosting-provider abuse report.

Correct affected data (measured 2026-09-08): >= 0.4.0, <= 0.7.2, no patched version.
Note this differs from NVD, which stops at 0.7.1 — v0.7.2 ships the identical vulnerable
MessageFilters.java (98 lines, GroovyScriptEngineImpl, case GROOVY_SCRIPT) and was
released 2024-04-10, after the CVE. :latest and :v0.7.2 share the digest
sha256:8f2ff02d64b0a7a2b71b6b3b3148b85f66d00ec20ad40c30bdcd415d46d31818.

The fix exists only in the fork kafbat/kafka-ui (commit 11a57d14, 2024-02-10, released
in v1.0.0), which renamed the package namespace from com.provectus.kafka.ui to
io.kafbat.ui — a migration to a different artifact, not an upgrade path.

Two asks:

1. Carry the CVE as an unreviewed advisory so it is at least visible and searchable in the
   database.
2. Consider this a data point for container-image coverage. I understand OCI images are
   outside the 12 supported ecosystems today. This case shows the gap has a concrete
   consequence: an image with 447M pulls, a known 8.8 RCE, an unmaintained upstream, and
   no tooling anywhere that flags it.

I have reported the empty OSV entry separately (google/osv.dev#5988), and a CVE update
request to MITRE to populate the structured affected fields is being submitted in
parallel — that record is the upstream cause, since it was published with
vendor/product/version all set to n/a.

Happy to open a PR here with advisory data in whichever schema you prefer, if that is
useful.