Kuberhealthy runs synthetic checks as Kubernetes pods and reports their health through a status UI, JSON API, and Prometheus metrics. The system is designed to reveal operational state, so the question is who can post results and who can read them.
On November 23, 2026, fixed count-only ZoomEye queries used sub_type=all, page 1, and page size 1. app="Kuberhealthy" returned 0 matches, title="Kuberhealthy" returned 2, and http.body="Kuberhealthy" returned 42.
The small internet footprint fits a cluster operator. A public product page is not the normal deployment model. The body matches may be documentation or status pages, and none of the queries tested the reporting endpoint.
The reporting credential
Kuberhealthy's check-creation documentation says each check pod receives KH_REPORTING_URL and KH_RUN_UUID. The run UUID must be sent in the kh-run-uuid request header to authenticate the report. The check has a deadline and posts a JSON object with an ok field and an error list.
That design gives a check pod a narrow capability: report the outcome of one run. The UUID should not be reused as a general service credential. Protect it as a secret and ensure logs do not expose the full reporting URL or header value.
If an attacker can forge a successful report, monitoring can hide a failure. If an attacker can mark a healthy service as failing, the team may waste time on a false incident. Both results affect the integrity of operational decisions.
Inventory the exposed views
Identify the status UI, JSON endpoint, metrics endpoint, and reporting endpoint. Confirm their network exposure and authentication. A read-only status page can still reveal service names, check cadence, and failure patterns.
Restrict metrics and JSON status to the monitoring network unless a public status page is intentional. Review the Kubernetes permissions of check pods and the controller. A check that only needs to report should not be able to modify unrelated resources.
Verify the signal
Test that a report with a missing or wrong run UUID is rejected. Confirm that expired runs cannot report late as success. Log accepted and rejected reports with the check identity and timestamp, but redact the UUID.
Review which teams can create checks and which images they may run. A custom check is executable code in the cluster. Use image allowlists or admission policy where the deployment supports them.
The internet count is almost negligible, which is why it should not be used as a coverage claim. Kuberhealthy's security value is in the integrity of the reporting path, the exposure of status data, and the authority of the check pods.