It was discovered that containerd incorrectly handled HTTP/2 SETTINGS frames. A remote attacker could possibly use this issue to cause containerd to enter an infinite loop, resulting in a denial of service. (CVE-2026-33814) Jakub Ciolek and Kyle Elliott discovered that containerd incorrectly handled group parsing when creating containers from images. An attacker could possibly use this issue to cause containerd to consume excessive memory, resulting in a denial of service. (CVE-2026-47262) Henry Beberman and Robert Prast discovered that containerd incorrectly validated image references when importing container checkpoints. An attacker could possibly use this issue to poison the local image cache and execute arbitrary code in other pods. (CVE-2026-50195) Robert Prast discovered that containerd incorrectly propagated labels from image configurations to containers. An attacker could possibly use this issue to execute arbitrary code on the host. (CVE-2026-53488) Yuming Zhang, Song Li, Sangwon Ryu, Henry Beberman, Robert Prast, Kyle Elliott and Zhenchen Wang discovered that containerd incorrectly validated symlinked paths when restoring container checkpoints. An attacker could possibly use this issue to read arbitrary files on the host, resulting in information disclosure. (CVE-2026-53489) Robert Prast discovered that containerd incorrectly trusted device interface annotations when restoring container checkpoints. An attacker could possibly use this issue to bypass resource allocation restrictions and inject devices or host mounts into a container. (CVE-2026-53492)
Continue reading...
Continue reading...
