408,609 Proxmox Fingerprints and 6 Hosts on 8006: When a Query Result Describes the Scanner
Two numbers from the same index, about the same product, look like they cannot both be right. One says 408,609. The other says 6. The explanation is not a data error, and working through it is the most useful thing this dataset offers.
Context and method
Counts from the ZoomEye index, 23 October 2026, sub_type=all, pagesize=1, page 1. Fingerprint matches, not installations.
| Query | Total |
|---|---|
app="Proxmox" |
408,609 |
title="Proxmox" |
212,739 |
app="Proxmox" && port=8006 |
6 |
Proxmox VE serves its management interface on 8006 by default. A query for the product fingerprint restricted to that port returns six hosts. The unrestricted fingerprint returns 408,609, and the title query returns 212,739.
Three readings, and only one of them is right
The first reading is that Proxmox management interfaces are almost never exposed, which would make the 408,609 figure refer to something else entirely.
The second is that the index has not finished scanning 8006, so the port-scoped query measures coverage rather than exposure.
The third is that app="Proxmox" matches a much broader set of hosts than Proxmox VE management interfaces: virtual machines running a Proxmox-branded service, backup tooling, documentation, reseller pages, and hosts mentioning the name in a header the fingerprint rule accepts.
Without an authenticated view of the underlying data, none of these can be ruled out. What can be said is that the port-scoped query and the fingerprint query measure different populations, and the 408,000-host difference is a caution against reading either as an exposure count.
What the numbers do support
They support a statement about fingerprint breadth. A name-based application rule matching 408,609 hosts while the product's default management port matches 6 is a rule with a very wide aperture. That matters when the index is used for inventory: a large number in a product's app dimension is not automatically a large exposed population, and the port-scoped variant is the cheapest way to test whether the rule is precise.
The right question for a hypervisor
Proxmox VE is a hypervisor management plane. One exposed console is enough to matter, because the console grants control over every guest on the host, and the guests usually include the systems that do everything else. The exposure question is therefore never about aggregate counts.
- Is the management interface reachable from anything other than an administrative network?
- Is it behind a second authentication factor, or only a password?
- Are the API tokens issued to automation scoped to a role rather than to root-equivalent permissions?
- Do the guests inherit credentials from the host, such as a hypervisor-level backup account?
Those answers are properties of your deployment. No external index, however large, substitutes for them.
A verification sequence
- Query
app="Proxmox"scoped to your own netblocks, then query the port-8006 variant. The difference between the two counts inside your own address space is a measurement of fingerprint breadth you can actually interpret. - For each host answering on 8006, verify the certificate and the login page directly, then stop. A management console is not something to enumerate automatically.
- Confirm the console is unreachable from the internet unless it is deliberately published, and if it is published, confirm what sits in front of it.
- Audit API tokens and their role assignments, because token abuse does not require the console at all.
What this piece is really about
6 and 408,609 came from the same API within seconds of each other. Anyone can query this index; almost nobody reconciles two results before publishing one of them. The reconciliation is where the value sits, and a number that survives it is worth more than a number that merely looks large.
References
- ZoomEye search: https://www.zoomeye.org/
- Proxmox VE administration guide: https://pve.proxmox.com/pve-docs/











