CVE-2026-44939: Rancher vulnerable to command injection through unsanitized YAML parameter

A critical command injection vulnerability has been identified in the Rancher Manager cluster import endpoint /v3/import/{token}_{clusterId}.yaml through unsanitized YAML parameters. This endpoint accepts an authImage query parameter that is rendered without sanitization into a generated Kubernetes manifest template. By including URL-encoded newlines in the parameter value, an attacker can break out of the image: field to inject arbitrary YAML keys and malicious configurations, such as commands to execute malicious containers.

Exploitation of this vulnerability requires the following conditions to be met:

• Attackers must obtain a valid cluster registration token (these tokens may be exposed, for example, through documentation, screenshots, or insecure communication channels).
• The victim’s cluster operator must execute kubectl apply against a maliciously crafted URL.

When a victim applies this compromised manifest using kubectl apply, a DaemonSet is deployed with the injected configuration. This DaemonSet:

• Runs on all control-plane nodes with hostNetwork: true enabled.
• Uses the cattle service account, which possesses cluster-admin privileges.
• Mounts /etc/kubernetes directly from the host.
• Executes attacker-controlled commands via the injected command: field.

An attacker who successfully exploits this vulnerability could:

• Achieve full control over downstream Kubernetes clusters.
• Execute arbitrary code on control-plane nodes with elevated privileges.
• Access sensitive cluster secrets and configurations via the privileged service account.
• Disrupt cluster operations by manipulating critical control-plane workloads.
• Establish persistent access through the deployed DaemonSet.

Note: If you believe that you might have been impacted by this vulnerability, it’s highly advised to review your clusters’ logs and deployment logs for signs of malicious deployments and to rotate all service accounts and credentials that might have been exposed in such a scenario.

Please refer to the associated MITRE ATT&CK - Technique - Deploy Container for further information about this category of attack.