Perses Defaults Make Authentication and the Datasource Proxy One Review
Perses is an observability dashboard server, but its security-relevant behavior reaches beyond rendering charts. It can proxy requests to datasources, and the current security documentation states that authentication is disabled by default. A reachable Perses API may therefore expose both dashboard management and a path from the server to internal services.
What the measurement found
On 2026-11-19, a count-only ZoomEye SDK query used sub_type=all, page 1, and pagesize 1. app="Perses" returned zero. title="Perses" returned 51, and http.body="Perses" returned 479.
The low title count is not proof of safety. The body count may include documentation, client assets, and error pages. Conversely, a real Perses deployment may not expose a distinctive title at all. Product identification must be paired with an application response and an owner record.
Understand the default risk
Perses documentation states that security.enable_auth is false in the shipped configuration. In that mode, anyone who can reach the API can create, modify, or delete projects, dashboards, datasources, and secrets. The datasource proxy can also make the server issue requests to any reachable URL and return the response.
The configuration documentation expands the consequence: a user allowed to create a datasource or use unsaved proxy endpoints may reach services available from the Perses server, including cloud metadata endpoints or the Kubernetes API. That is a server-side request capability, not a dashboard-display issue.
Review in two layers
The first layer is authentication and authorization. Confirm security.enable_auth, the provider configuration, the encryption key source, and the role model. Check whether anonymous access is intentionally allowed and which actions it grants. If authentication is enabled without a proper encryption key or provider, test the complete login and secret-storage path rather than assuming the setting is sufficient.
The second layer is the datasource proxy. Inventory global, project, and local datasources. For each one, record the target URL or database, authenticated client, owner, scope, and purpose. Restrict datasource creation to a small role. Review whether the proxy can reach metadata services, control planes, or unrelated internal networks.
SQL proxy mode should also be examined. Documentation describes read-only transaction behavior, but read-only does not mean harmless. A database query can still disclose sensitive records or reveal service topology. Apply database-level least privilege and network controls in addition to the proxy setting.
Validate a candidate
Use title and body leads, then confirm through an approved response and ownership record. Do not call resource-changing endpoints or test the proxy against third-party systems. An owner can identify the deployment and confirm its configuration.
If the service is a sandbox or an internal evaluation instance, verify that it is not reachable from untrusted networks. If it is production, review authentication, secrets, datasource scope, and network egress together.
The measured counts are small compared with broad infrastructure products, but size is not the decision. A single unauthenticated Perses instance with a powerful datasource proxy can provide more access than a much larger read-only dashboard inventory.
References
- Perses security considerations, Perses documentation. Retrieved 2026-11-18.
- Perses proxy, Perses documentation. Retrieved 2026-11-18.
- Perses configuration, Perses documentation. Retrieved 2026-11-18.
- ZoomEye counts:
app="Perses"0;title="Perses"51;http.body="Perses"479.sub_type=all, page 1, pagesize 1, observed 2026-11-19 UTC+8.
Top comments (0)