Improper kubeconfig validation allows arbitrary code execution

Flux2 can reconcile the state of a remote cluster when provided with a kubeconfig with the correct access rights. Kubeconfig files can define commands to be executed to generate on-demand authentication tokens. A malicious user with write access to a Flux source or direct access to the target cluster, could craft a kubeconfig to execute arbitrary code inside the controller’s container.

In multi-tenancy deployments this can also lead to privilege escalation if the controller's service account has elevated permissions.

Impact

Within the affected versions range, one of the permissions set below would be required for the vulnerability to be exploited:

• Direct access to the cluster to create Flux Kustomization or HelmRelease objects and Kubernetes Secrets.
• Direct access to the cluster to modify existing Kubernetes secrets being used as kubeconfig in existing Flux Kustomization or HelmRelease objects.
• Direct access to the cluster to modify existing Flux Kustomization or HelmRelease objects and access to create or modify existing Kubernetes secrets.
• Access rights to make changes to a configured Flux Source (i.e. Git repository).

Patches

This vulnerability was fixed in kustomize-controller v0.23.0 and helm-controller v0.19.0, both included in flux2 v0.29.0. Starting from the fixed versions, both controllers disable the use of command execution from kubeconfig files by default, users have to opt-in by adding the flag --insecure-kubeconfig-exec to the controller’s command arguments. Users are no longer allowed to refer to files in the controller’s filesystem in the kubeconfig files provided for the remote apply feature.

Workarounds

• The functionality can be disabled via Validating Admission webhooks (e.g. OPA Gatekeeper, Kyverno) by restricting users from being able to set the spec.kubeConfig field in Flux and objects.