DEV Community

yutianle
yutianle

Posted on

Designing a repeatable internet-measurement record for audits

Designing a repeatable internet-measurement record for audits

Security teams often cite a search count without preserving enough information to reproduce it. That becomes a problem when a number is used in an incident review, a board paper, or a follow-up investigation. The ASD's cPanel and SonicWall cases provide a useful test: can an analyst show exactly what was measured, when, and what the result did not prove?

The ASD report says a collective cPanel notification reached 340 entities in May 2026 and was followed by more than 60 confirmed incidents. In the SonicWall case, the agency identified almost 1,000 potentially vulnerable devices across around 40 organisations. Those numbers are contextual evidence from a public authority. A ZoomEye measurement is a separate observation of public assets and must be labeled that way.

Record the query as evidence

A repeatable record should include the exact dork, subtype, collection time, count, status, and deterministic search link. For this event, the cPanel query app="cPanel" returned 10,117,122 all-scope observations at 14:41 UTC on 17 November 2026. The Australia-scoped query returned 224,115. The SonicWall query returned 2,518,524 globally and 17,878 in Australia. The CVE-2024-40766 query returned 93,984, while CVE-2026-41940 returned zero.

The cPanel search link is ZoomEye cPanel results. The SonicWall link is ZoomEye SonicWall results. Store these links with the original query text. A screenshot alone is weaker because it can omit the scope and timestamp.

Write a result statement that survives review

Use a sentence that names the unit and the limit. For example: "At 14:41 UTC on 17 November 2026, ZoomEye reported 10,117,122 all-scope public cPanel product observations for app=\"cPanel\". This is a product-exposure baseline, not an affected-instance or victim count."

Repeat the structure for every query. Avoid combining results from different timestamps. Avoid adding global and regional counts. Avoid treating a zero CVE-index count as proof that no vulnerable systems exist. The exact CVE search for CVE-2026-41940 returned zero while the product query returned 10,117,122, which is a clear example of why the fields answer different questions.

Keep an evidence package

The audit folder should contain:

  1. The source article, alert, and vendor advisory with retrieval dates.

  2. The ZoomEye query and the raw command result or stored evidence ID.

  3. The normalized dork, subtype, total count, and UTC timestamp.

  4. The definition of the unit, including whether it is a product, CVE, domain, or network observation.

  5. The local asset match, owner, build check, and remediation record.

  6. A note on limitations, including any failed query, zero index result, or attribution gap.

This package can be reused during a later alert. It also lets a reviewer distinguish a technical measurement from a claim in the source article. The ASD report's 340 entities and 60 confirmed incidents remain authority-reported facts. The ZoomEye total remains a platform observation.

Set review points

Re-run a baseline only when there is a decision to make. A scheduled comparison can support long-term monitoring, but every result should have an owner and an intended use. Before closing an item, check that the quoted number still has its original query, time, and scope. Before escalating an item, check that the candidate has been mapped to an owned asset and that the product or CVE actually matches the risk being discussed.

The process is deliberately plain. It keeps a public measurement from drifting into an incident claim and gives the next analyst enough evidence to continue. For a cPanel or SonicWall review, that is the minimum standard for an auditable conclusion.

References

Top comments (0)