This vulnerability arises from CRI-O's checkpoint restore implementation restoring process credentials, Linux capabilities, the no_new_privs bit, and seccomp state directly from checkpoint data instead of enforcing the Kubernetes pod's security context. An attacker with permission to create pods can craft malicious checkpointed container images that execute with root privileges and full capabilities, bypassing security policies. The checkpoint restore from remote registries is enabled by default in affected versions, making exploitation possible without local node access. Affected upstream versions include CRI-O 1.34 and later; downstream Red Hat OpenShift Container Platform versions 4.17 and later are also impacted. Fixes have been developed for CRI-O branches 1.34.x, 1.35.x, and 1.36.x but are not yet released. Until patches are available, mitigation includes disabling checkpoint restore if not required, restricting pod creation privileges, enforcing Pod Security Standards, and restricting container image registries to trusted sources.

Successful exploitation allows an attacker with pod creation privileges to execute processes with elevated privileges (root and full Linux capabilities) inside containers, bypassing Kubernetes security context enforcement. This can lead to unauthorized code execution, privilege escalation, and potential compromise of containerized workloads. The vulnerability affects confidentiality, integrity, and availability of affected systems. Exploitation requires checkpoint restore functionality to be enabled and the ability to create pods from malicious checkpoint images.

A fix is available but not yet released; updates will be provided for CRI-O versions 1.34.14, 1.35.9, and 1.36.6, and corresponding OpenShift Container Platform updates. Until patches are released, mitigate by disabling checkpoint restore if not needed (disable ContainerCheckpoint feature gate in OpenShift or set enable_criu_support=false in crio.conf for standalone CRI-O). Restrict Kubernetes RBAC permissions to limit pod creation to trusted users only. Enforce Pod Security Standards at the namespace level to limit security contexts. Restrict container image registries to trusted sources using image policies. Identify and remediate containers restored from untrusted checkpoint images by stopping, deleting, and recreating them. Monitor Red Hat advisories for release of official fixes.