EUVD-2026-35499
GHSA-x5qp-7hqq-hm2j
Severity by source
AV:L/AC:L/PR:L/UI:R/S:U/C:H/I:H/A:H
Primary rating from Vendor (nvidia) · only source for this CVE.
CVSS VectorVendor: nvidia
Lifecycle Timeline
1DescriptionCVE.org
NVIDIA DALI contains a vulnerability in a component where an attacker could cause a heap-based buffer overflow. A successful exploit of this vulnerability might lead to code execution, data tampering, denial of service, and information disclosure.
AnalysisAI
Heap-based buffer overflow in NVIDIA DALI enables local authenticated attackers to achieve code execution, data tampering, denial of service, or information disclosure when a victim user interacts with attacker-supplied input. The flaw affects the Data Loading Library used in GPU-accelerated deep learning data pipelines and carries a CVSS 3.1 score of 7.3 (High). No public exploit identified at time of analysis and the issue is not listed in CISA KEV.
Technical ContextAI
NVIDIA DALI (Data Loading Library) is a GPU-accelerated library for decoding and augmenting image, video, and audio data feeding deep learning frameworks such as PyTorch, TensorFlow, MXNet, and PaddlePaddle, identified by CPE cpe:2.3:a:nvidia:dali. The root cause is CWE-122 (Heap-based Buffer Overflow), a memory-safety class typically arising in native C/C++ parsing or decoding routines where bounds checks on attacker-controlled length or offset fields are missing or incorrect, allowing writes past the end of a heap-allocated buffer and corruption of adjacent heap metadata or object pointers. In a library that ingests media files for ML training and inference, such overflows commonly occur in format-specific decoders or batched tensor construction paths processing untrusted sample data.
RemediationAI
Patch available per vendor advisory - consult the NVIDIA security bulletin at https://nvidia.custhelp.com/app/answers/detail/a_id/5814 for the exact fixed DALI version and upgrade the library across all training and inference nodes (pip install --upgrade nvidia-dali-cudaXXX or the equivalent conda/container image refresh), then rebuild any container images, MLOps pipelines, or model-serving artifacts that bundle DALI. Until the patched release is rolled out, restrict DALI to trusted local users only, avoid feeding it untrusted datasets or media files from external contributors, and treat shared multi-tenant GPU notebook environments as elevated-risk; isolating training jobs in per-user containers or VMs limits the blast radius of a successful overflow at the cost of additional orchestration overhead. Reference NVD (https://nvd.nist.gov/vuln/detail/CVE-2026-24180) and CVE.org (https://www.cve.org/CVERecord?id=CVE-2026-24180) for tracking.
Arbitrary code execution in NVIDIA DALI (all versions prior to 2.0) allows local authenticated attackers with low privil
Local code execution and data tampering in NVIDIA DALI (Data Loading Library) is possible when a low-privileged user is
Same weakness CWE-122 – Heap-based Buffer Overflow
View allShare
External POC / Exploit Code
Leaving vuln.today