Privilege Escalation Vulnerability in NVIDIA GPU Kernel Mode Driver for Linux
Overview
A high-severity vulnerability was disclosed in the NVIDIA GPU kernel mode driver, a critical component for AI/ML infrastructure. The flaw allowed a local, unprivileged user to escalate their privileges to root on the host machine. The vulnerability stemmed from a race condition in how the driver managed memory mappings for user-space clients. A malicious user could craft a sequence of IOCTL calls to the `/dev/nvidia` device file that would trigger a use-after-free condition in the kernel. By carefully grooming kernel memory, an attacker could replace a freed object with a controlled payload, leading to arbitrary code execution within the kernel context. This is particularly dangerous in multi-tenant GPU-accelerated Kubernetes clusters, where different users run containerized workloads on shared nodes. An attacker could use this vulnerability to break out of their container, gain full control of the underlying host node, and subsequently access or disrupt all other ML training jobs and data on that machine. The issue required local access but was exploitable from within a standard Docker container with GPU access.
Affected Systems
Testing Guide
1. **Check Driver Version**: Log into the GPU host machine and run the command `nvidia-smi`. 2. **Inspect Output**: Look at the 'Driver Version' displayed in the top right corner of the output. 3. **Compare Versions**: If the reported driver version is less than `535.129.03`, the system is vulnerable and must be patched.
Mitigation Steps
1. **Update Drivers**: Immediately update all NVIDIA drivers on host machines to version `535.129.03` or later. 2. **Restrict GPU Access**: Limit which users and pods can request GPU resources in a shared cluster environment. 3. **Use Sandboxed Runtimes**: Employ container runtimes with stronger kernel isolation, such as gVisor or Kata Containers, for running untrusted GPU workloads. 4. **Monitor Host Integrity**: Use host-based intrusion detection systems (HIDS) to monitor for suspicious activity or unauthorized privilege escalation on GPU nodes.
Patch Details
Patched in NVIDIA driver version 535.129.03 and subsequent releases as part of the November 2025 NVIDIA security bulletin.