Tianji combines website analytics, uptime monitoring, and server-status collection in one self-hosted application. For defenders, those are not equivalent surfaces. A public analytics script can be harmless on a customer site, while a reachable monitoring server may hold agent credentials, incident history, and a path into the infrastructure it observes.

I measured two views of the product on ZoomEye with sub_type=all, page 1, and pagesize=1. The query title="Tianji" returned 665 index matches. The application query app="Tianji" returned one. The checks ran on 2026-11-22 at 00:45 UTC. A match is an asset in the search index, not proof that the service is vulnerable or that the organization owns the deployment.

The title result is useful as a discovery lead and weak as a product inventory. "Tianji" can appear in page titles, documentation, templates, or unrelated sites. The single application match is stronger evidence of a recognized deployment, but it is too small to estimate the installed base.

Why the two numbers diverge

ZoomEye's app field depends on a product fingerprint. A self-hosted app can be customized, hidden behind a reverse proxy, or return a title and assets that do not match the current fingerprint. Rebranding, an older release, or a white-label deployment can move a host from app to title. The opposite can also happen when a generic title is omitted from a page but the application still exposes a recognizable response.

The gap matters because Tianji joins collection and control in one interface. Website analytics records visitor events. Uptime checks reveal service names and internal endpoints. Server-status agents can report machine health and may carry a token or network path used to submit data. The control plane is the more sensitive target, even if the public analytics script has a much larger web footprint.

Review the title result in ZoomEye. Use it to build candidate names, then verify the actual product, hostname, and ownership. Do not send a warning to every organization whose page contains the word.

Build an inventory without confusing the roles

Start with the application match and validate the response headers, login page, asset paths, and API behavior against Tianji documentation. Then expand with title matches only when the page contains product-specific scripts, paths, or fingerprint material. A title match needs a second signal before it becomes a deployment record.

Classify each confirmed instance by role. A public analytics collector should have a narrow ingestion path and an authenticated dashboard. A monitoring hub should be reachable only from approved operators or a private network. A status worker should not expose its enrollment token in a public page.

Check for default accounts, exposed API documentation, agent tokens, and unaudited workspace membership. The product's monitoring value comes from its access to other systems, so the security review must follow that access rather than stop at the login page.

Monitor the signal that matters

Re-run the two queries as separate baselines. A rising title count with a flat application count may mean custom deployments or title noise. A change in the application count is stronger evidence of a new or removed fingerprint. Neither movement proves compromise.

For an organization-owned host, compare the ZoomEye observation with DNS, TLS certificate ownership, deployment records, and the monitoring agent inventory. The external evidence is a lead. Internal logs and configuration establish whether the service is expected.

The most useful first step is to find the one application-fingerprinted instance and explain why it is visible. The 665 title matches show how a product name spreads across the web. The one app match is the narrower, more actionable signal.

References