Building an external asset register for supply-chain response
Incident response often starts with a list of known systems. The testmetest incident shows why that list needs an external companion. The campaign involved 42 GitHub repositories across 11 owners, npm publication infrastructure, and shellpanel workflows linked to cPanel scanning. An internal CMDB may know none of those external observations until someone asks the right question.
An external asset register is a controlled record of what the internet view showed, when it showed it, and how the observation was validated. It is not a victim list.
Why a separate register helps
ZoomEye and internal inventory answer different questions. ZoomEye can show that a public product fingerprint exists. Internal inventory can show who owns the address, which business service uses it, and whether the service is approved for public access.
Keeping the observations separate prevents a global count from being copied into an incident record as an affected count. It also preserves the evidence needed to explain why a host entered the review queue.
A record for each observation
A useful record contains:
| Field | Purpose |
|---|---|
| asset_id | Stable internal identifier after ownership is confirmed |
| observed_address | IP, domain, or product observation reference |
| product | cPanel, Jenkins, GitLab, Artifactory, or another observed service |
| dork | Exact normalized search expression |
| collected_at | Time of the observation |
| scope | Subtype, geography, ownership, and filters |
| count_basis | Whether the row is a global count or an owned subset |
| owner | Team or provider responsible for validation |
| patch_state | Confirmed version or unknown |
| access_review | Authentication-log outcome |
| evidence_links | Source record, scan result, and host evidence |
| status | Candidate, validated, remediated, or unresolved |
A product row alone should not set status to affected. The status moves forward only when ownership, access, or host evidence supports it.
Observations from this event
On November 19, 2026, stored ZoomEye queries returned 10,122,751 public app="cPanel" observations, 36,962 app="Jenkins", 1,297,214 app="GitLab", and 3,886 app="JFrog Artifactory" observations. These are global product-fingerprint baselines.
The event evidence adds context without changing the meaning of those counts. JFrog found armed repositories and a hidden CI publisher. The same owner cluster contained shellpanel workflows that targeted exposed cPanel systems through CVE-2026-41940. cPanel states that the affected range began after v11.40 and includes WP Squared through v11.136.1.6.
The register should therefore hold repository records, package and runner records, and external administration-plane records as linked but separate entities.
Ownership resolution
Resolve ownership before escalating a public observation. Use DNS, certificate data, address registration, hosting contracts, cloud inventory, and application owners. A hosting provider may own the address while a customer owns the panel. Record both relationships when they exist.
For an unowned observation, change the validation status instead of guessing. Unowned public infrastructure can remain outside the incident's responsibility while still mattering to the network provider that operates it.
Preserve query lineage
A count without the dork, scope, and time cannot be reused safely. Store the query next to the result. If the query changes from product fingerprint to CVE index or change filter, create a new observation row. Do not overwrite the earlier evidence.
This is especially important for zero results. A successful zero for a CVE field can document coverage at one time, but it cannot replace the product record or become proof that no host was vulnerable.
Maintain the register
Run scheduled external observations with stable scopes. Refresh ownership links, patch state, and authentication review separately. Compare new observations with the previous baseline and show the evidence difference in the incident view.
The register is useful when it lets responders answer three questions without mixing them: what was observed, who owns it, and what did the host evidence prove?
