Dockhand EUVD-2026-102920

Exploitation requires: (1) network reachability to a Dockhand instance running a version prior to 1.0.34; (2) knowledge or enumeration of a valid, running target container ID (the endpoint is /api/containers/<id>/exec); and (3) a target container that exposes an interactive shell. No authentication is needed because the WebSocket upgrade enforces none (CWE-306, PR:N). The command execution lands inside the container; full host compromise is a conditional escalation that applies ONLY when the targeted container runs in privileged mode or has sensitive host mounts (e.g., the Docker socket or host filesystem paths). The practical limiting factors for defenders are the need to know a specific container ID and the requirement that the container have an interactive shell - these constrain blind mass exploitation even though no credentials are required.

This is a genuine high-priority issue, not a CVSS-inflated one: CWE-306 (missing authentication) on the /api/containers/<id>/exec WebSocket upgrade means an unauthenticated actor who can reach Dockhand can open an interactive shell inside a container, and the S:C/C:H/I:H/A:H vector reflects realistic container-to-host escalation when the target container is privileged or has sensitive mounts (e.g., the Docker socket). The input provides NO EPSS score, NO CISA KEV listing, and NO public POC, so there is no public exploit identified at time of analysis and no evidence of active exploitation - the 9.6 severity is driven purely by the technical vector, not by observed attacker interest. Note a conflict between signals worth verifying: the vendor-supplied vector uses AV:A (adjacent), whereas the description ('has network reach to Dockhand') implies a general network-reachable WebSocket endpoint (AV:N), which would raise real exposure for instances exposed beyond a management segment - verify Dockhand's typical network placement. Two practical limiters temper the raw score: the attacker must know a valid target container ID and the target container must expose an interactive shell, so this is realistically exploitable but requires some target knowledge rather than being a blind internet-wide mass-exploitation primitive.

An attacker with network access to a Dockhand instance older than 1.0.34 identifies or enumerates a running container ID and opens a raw WebSocket to /api/containers/<id>/exec, which performs no authentication on the upgrade. They obtain an interactive shell and run arbitrary commands inside that container; if the container is privileged or mounts sensitive host paths (such as the Docker socket), they pivot to full host compromise. No public exploit was identified at time of analysis, and the low attack complexity of the auth flaw itself makes a working client trivial to build once a container ID is known.