CVE-2024-0132 is a critical time-of-check time-of-use vulnerability in NVIDIA Container Toolkit 1.16.1 and earlier on Linux when operated in its default configuration. The flaw arises during initialization while the toolkit processes container-provided library paths and does not safely resolve symbolic links across the container-host boundary. A specially crafted container image can exploit this race condition and symlink handling weakness to cause privileged host-side toolkit components to access unintended locations on the host filesystem. This can break container isolation and enable host filesystem access from a container context. Deployments using Container Device Interface (CDI) are not affected. NVIDIA GPU Operator deployments that include affected toolkit versions are also impacted.
Mallory correlates every CVE against your assets, your vendors, and active adversary campaigns. Know which vulnerabilities matter for you, not just which ones are loud.
What it means. What to do now. Patch path, mitigations, and the assume-compromise checklist.
What an attacker gets, and what they’ve been doing with it.
If you can’t patch tonight, do this now.
Patch, then assume compromise.
2 valid exploits after Mallory filtered fakes, detection scripts, and README-only repos (2 hidden).
This repository is a proof-of-concept (PoC) exploit for CVE-2024-0132, targeting NVIDIA's nvidia-container-toolkit on Linux systems. The exploit demonstrates how a malicious container, when run with the NVIDIA runtime and GPU access, can escape its isolation and access files from the host filesystem and interact with the host's Docker daemon via the docker.sock UNIX socket. The repository contains multiple versions (v0.1, v0.2, v0.3), each with a Dockerfile and an entrypoint.sh script. The exploit works by creating symlinks within the container that point to host directories, then using these to read sensitive files (e.g., /etc/hostname) and communicate with the Docker daemon. The main payload is a bash script executed as the container's entrypoint. The PoC is operational and demonstrates real host compromise potential, but does not include weaponized or automated post-exploitation features. The README provides detailed setup and reproduction steps, including environment requirements and affected versions.
This repository provides a proof-of-concept (PoC) exploit for CVE-2024-0132, a TOCTOU (Time-of-Check to Time-of-Use) vulnerability in the NVIDIA container toolkit (libnvidia-container) on Linux systems. The exploit is implemented as a Dockerfile that sets up a specific directory and symlink structure to abuse the way the toolkit mounts libraries from /usr/local/cuda/compat/ into the container. By carefully crafting directories and symlinks, the PoC demonstrates how an attacker could trick the mounting logic into mounting arbitrary host files or directories into the container, potentially leading to privilege escalation or access to sensitive host data. The repository contains two files: a Dockerfile with the exploit logic and a README.md that explains the vulnerability, the exploit steps, and references to the relevant code in the NVIDIA toolkit. The exploit is local in nature, requiring the ability to build and run Docker containers with the NVIDIA toolkit.
Products and vendors Mallory has correlated with this vulnerability. Open in Mallory to drill down to specific CPE configurations and version ranges.
Vendor-confirmed product mapping. Mallory continuously reconciles this list against your asset inventory.
19 sources tracked across advisories, community write-ups, and news. New activity surfaces here as Mallory finds it.
Критическая TOCTOU/path traversal container escape уязвимость в NVIDIA Container Toolkit, позволяющая через подмену symlink заставить привилегированный хостовый процесс обращаться к файловой системе хоста.
An NVIDIA Container Toolkit TOCTOU vulnerability that can enable container escape; discussed in context of an incomplete patch leaving exposure.
A container escape vulnerability in the NVIDIA Container Toolkit mentioned as part of NVIDIA's broader security history.
A previously addressed NVIDIA AI infrastructure vulnerability mentioned as part of vendor security history; specific technical details are not provided in the content.
Query your assets running an affected version, and investigate the blast radius.
Every observed campaign linking this CVE to a named adversary.
Malware families riding this exploit, with evidence and IOCs.
YARA, Sigma, Snort, and vendor rules, auto-deployed to your SIEM.
Cross-references every affected SKU, including bundled OEM variants.
Community discussion across Reddit, Mastodon, and other social sources.