Quick wins for a faster PC:
Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →Repair Windows errors before they cause bigger problemsFix Now →Scan for outdated or missing drivers - takes under a minuteDriver Scan →In containerd’s affected CRI checkpoint-restore path, a restored process can resume with security attributes saved in its checkpoint instead of receiving the restrictive settings requested by the destination Pod or CRI configuration. The gap is specific to Linux containerd CRI checkpoint restore: ordinary container creation is not the vulnerable route described in the advisory.
What failed: requested settings were not the restored process’s effective settings
A destination configuration can request controls such as restricted credentials, Linux capabilities, no_new_privs, and seccomp. But when containerd restores a process from checkpoint data through the affected CRI CreateContainer path, the containerd project says CRIU restores those attributes from the checkpoint rather than enforcing the destination CRI ContainerConfig.
As an Amazon Associate I earn from qualifying purchases.
That means a crafted checkpoint image could cause a restored process to run as root with full capabilities and without the requested seccomp filters, even when the destination configuration asks for tighter security. The security boundary is the restore operation: the checkpoint is an input to process reconstruction, and its saved process state can take precedence over destination policy.
Free tools Windows power users keep installed
One-click scans. No signup required.
This is not a claim that every checkpoint restore is malicious or that every Pod creation is vulnerable. The containerd advisory rates the issue Critical, but that severity does not establish how many clusters are exposed or how often the affected path is used.
#1 Best Overall
Which systems and workloads are in scope?
The affected conditions are specific: the system is Linux, containerd’s CRI checkpoint restore is in use, and an attacker can run a container using a crafted checkpoint image. The containerd advisory says users who do not use CRI checkpoint restore are not affected. Google also says standard container creation in GKE Standard and GKE Autopilot remains unaffected.
- Potentially affected: a workload restored through the affected CRI path from an attacker-controlled or otherwise untrusted checkpoint image, on an affected containerd version.
- Not the described route: ordinary creation of a new container without CRI checkpoint restore.
- Not established: the prevalence of checkpoint restore or the number of exposed clusters. The advisory’s Critical rating is a severity assessment, not an exposure count.
Why status can look correct while process state is not
Containerd reports the requested configuration in CRI status; according to its advisory, that status does not report the restored process’s actual security state. As a result, a control-plane view can show the intended policy without proving that the process is running with it.
For an operational check, identify the restored process’s PID on the node and inspect its Linux process status, then compare the values with the destination Pod’s requested security context:
Recommended Free Tools
grep -E 'NoNewPrivs|Seccomp|CapEff' /proc/<pid>/statusThe command reads the process’s NoNewPrivs, Seccomp, and effective-capabilities fields. It is a diagnostic of process state, not a substitute for upgrading or disabling the affected restore path. A CRI status record alone cannot establish that the restored process matches the requested policy.
Which containerd versions are affected, and what changes?
The containerd advisory lists these affected ranges and remediation points:
| Containerd version or range | Checkpoint restore via CRI CreateContainer |
Operator implication |
|---|---|---|
>= 2.1.0 < 2.2.7 |
Affected range | Update to a fixed release; the advisory says there is no configuration option to disable this restore path on unpatched versions. |
2.2.7 |
Disabled by default | Do not re-enable restore via CreateContainer unless the risk is addressed; the advisory says re-enabling leaves the vulnerability present. |
>= 2.3.0 < 2.3.4 |
Affected range | Update to a fixed release; the advisory says there is no configuration option to disable this restore path on unpatched versions. |
2.3.4 |
Disabled by default | Do not re-enable restore via CreateContainer unless the risk is addressed; re-enabling leaves the vulnerability present. |
2.4.0 |
Feature removed | The restore feature is no longer available through this path. |
These version statements are from the containerd project’s advisory. For a managed Kubernetes service, check the provider’s bulletin and the actual runtime version on the nodes; a Kubernetes control-plane version by itself does not establish which containerd build is installed.
What should an operator do?
- Determine whether the restore path is used. Check whether Linux nodes run containerd with CRI checkpoint restore enabled and whether workloads can be created from checkpoint images. If the path is not used, the advisory says the issue does not affect the environment.
- Check the installed containerd release. Compare it with the affected ranges above and follow the applicable containerd or cloud-provider update guidance. On unpatched versions, the advisory says there is no configuration option to disable this restore path.
- Do not restore untrusted checkpoints. The advisory recommends stopping and deleting containers restored from untrusted checkpoints, then recreating them rather than continuing to rely on their requested configuration.
- Limit who can create containers or Pods. Google recommends restricting container-creation permissions. Apply that control to the identities and workflows able to introduce checkpoint images or invoke the relevant restore path.
- Validate image registries and monitor node activity. Google recommends validating registries and watching node logs and runtime events for suspicious restore activity.
- Verify effective process state when needed. Inspect the restored process under
/procas described above and compare it with the intended security context. Do not treat the reported CRI configuration as proof of enforcement.
Related checkpoint risks are separate issues
Other containerd advisories show why a checkpoint should be treated as security-sensitive input, but they describe distinct failure modes and conditions—not additional effects that necessarily accompany this process-security issue:
- A June symlink-following issue could let a restored
container.logsymlink expose arbitrary host files throughkubectl logs. Its listed fixes are containerd 2.1.9, 2.2.5, and 2.3.2. - A CDI issue involved untrusted checkpoint metadata carrying CDI annotations into restoration, potentially bypassing normal Kubernetes resource allocation and device-plugin enforcement when CDI and matching host-spec conditions applied.
- A checkpoint-import issue involved unvalidated image references that could poison the node-local image cache.
These examples concern different assets and prerequisites. They reinforce the need to control checkpoint provenance and restore authority, but they do not mean every restore has all of these defects.
Best Value
What Kubernetes’ checkpoint proposal does—and does not—settle
Kubernetes KEP-5823 proposes Pod-level CheckpointPod and RestorePod CRI APIs, kubelet handling, and declarative restore through Pod configuration. The proposal treats runtime checkpoint contents and format as opaque to Kubernetes and owned by the runtime/checkpoint mechanism. It also discusses the privilege model introduced by namespaced checkpoint/restore API objects.
That design direction does not establish that restored process security attributes match destination policy. Nor does the proposal establish, by itself, that the APIs are currently shipping or available in a particular release. The destination-policy enforcement question remains a runtime security question.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
Do these 3 things before closing this tab:
1Repair Windows errors before they cause bigger problems2Scan for outdated or missing drivers - takes under a minute3Clear out junk files and repair common Windows errors
