Kanea installs a daemon that runs as root, holds every secret on the node, and can open a shell in any container on it. Vulnerability reports are taken seriously and are welcome.
Do not open a public issue. Report privately through GitHub Security Advisories:
If that is not available to you, email michael@essandoh.dev.
Please include what you need to make the finding actionable: affected version
(kanea version), the component (kanead, kanea-edge, the API, the dashboard,
the MCP server, the installer), reproduction steps, and what an attacker gains.
You should get an acknowledgement within 5 days and an assessment within 14 days. Please give a fix a reasonable window before disclosing publicly; if a report is valid and you would like credit in the advisory, say so.
The latest release is supported. Kanea is pre-1.0 and there are no backported security releases: fixes ship in the next tag.
In scope is anything that crosses a boundary the platform claims to hold:
- Authentication or authorisation bypass on any API, websocket or MCP route (every one of them is deny-by-default; a route that is not is a bug).
- Anything that reveals a secret value, whether over the API, the MCP server, the dashboard, the audit log, or a log line.
- Escaping a workload's isolation: capabilities, seccomp, namespaces, the cgroup ceiling, or a volume mount reaching outside its storage.
- Crossing a project boundary: a spec, a synced repository, or a webhook that reaches another project's services, secrets or network.
- Breaking the release chain: the installer, the checksums, the Sigstore signature, or the archive encryption and its restore path.
- SSRF out of the notification or ACME egress paths.
Out of scope:
- Anything requiring an operator to already hold the master key or a root shell on the node. That is the trust the design starts from.
- Findings in containerd, BuildKit or the kernel. Report those upstream; tell us if Kanea's defaults make them materially worse.
- Automated scanner output with no demonstrated impact.
docs/THREAT_MODEL.md documents the boundaries, the
adversaries considered, and the OWASP Top 10 mapping as built. Reading it first
will tell you which assumptions are load-bearing, and a report that breaks one
of them is exactly the report worth sending.