Advisories

Image config labels flow into container metadata and can trigger host-root command execution through a label-consuming containerd plugin.

ghsa / CVE / Writeup

Checkpoint import can assign an attacker-controlled digest to a local image tag, poisoning the node cache for later pods using that tag.

ghsa / CVE / Writeup

Checkpoint restore copies container.log without rejecting symlinks, allowing host-file reads to surface through kubectl logs.

ghsa / CVE / Writeup

Untrusted checkpoint metadata can smuggle CDI annotations and bypass Kubernetes device allocation on nodes with matching CDI specs.

ghsa / CVE / Writeup

Secret sync name collision enables cross-namespace L7 policy bypass

Name collisions between Secret and ConfigMap objects let a namespace-scoped attacker overwrite synchronized SDS objects and bypass L7 network policy enforcement.

ghsa / CVE

Tool-name shadowing across aggregated servers

Aggregating MCP servers into a flat namespace without per-server prefixes or collision detection lets an untrusted server shadow trusted tool names and intercept calls.

ghsa / CVE

Direct ACME Challenge resources can bypass Issuer DNS01 solver policy and use ClusterIssuer DNS credentials

Directly created ACME Challenge resources can supply attacker-controlled DNS01 solver config while cert-manager uses ClusterIssuer credentials.

ghsa / CVE