Node Kernel Bugs on GKE COS Nodes Enable Host Escalation
Linux kernel vulnerabilities on Container-Optimized OS (COS) nodes in Google Kubernetes Engine (GKE) allow attackers with local access to escalate privileges to root on the host node. This breaks the isolation boundary between containers and the underlying node, meaning a compromised pod can lead to full control over the node and all other pods running on it. Standard patching of workloads alone does not address this risk because the vulnerable kernel remains until the node OS image is updated.
Google published multiple security bulletins (GCP-2026-018, -017, -016, and others) detailing high-severity Linux kernel privilege escalation flaws affecting COS nodes. Updates also include patched kernel versions for Ubuntu nodes running in GKE clusters. These advisories highlight a recurring focus on node-level kernel vulnerabilities rather than isolated container or control-plane issues. The affected environments are primarily GKE clusters using COS or Ubuntu node images.
The risk is significant for multi-tenant clusters and environments running untrusted workloads, where a single compromised container can lead to host takeover. This cluster-wide exposure persists until node OS updates are applied. The advisories emphasize the need to verify that node OS and kernel patches are current to maintain the security boundary between workloads and the host.