It has been found that the permissions of the /etc/kubernetes/kubeconfig on an OpenShift node are too permissive. An attacker with access to a running container which mounts /etc/kubernetes or has local access to the node, may copy this kubeconfig file and potentially attempt to add their own rogue node to the OpenShift cluster.
Acknowledgments: Name: Juan Osorio Robles (Red Hat)
Upstream fix/work in progress: https://github.com/openshift/machine-config-operator/pull/2202
Statement: The kubeconfig file on the node is that of the bootstrap certificate. This means that to successfully exploit this vulnerability, the kubeconfig file must be taken and used within the first 24 hours of creation as it is created with a short expiration time. Otherwise, attempting to use the credentials after this time will result in the following: `Failed while requesting a signed certificate from the master: cannot create certificate signing request: Unauthorized` Further, if taken, the certificate signing request (CSR) of the joining rogue OpenShift node is not automatically approved by default. The node still needs to be approved through the use of the commands `oc get csr` and `oc adm certificate approve`.
closing old flaw bug