EUVD-2026-44487
GHSA-8jxq-crv3-7q3x
Severity by source
AV:L/AC:L/PR:N/UI:R/S:U/C:H/I:H/A:H
Local file-parsing flaw needing the victim to load crafted input, so AV:L and UI:R; no auth required (PR:N); code execution yields full C:H/I:H/A:H.
Primary rating from Vendor (nvidia).
CVSS VectorVendor: nvidia
Lifecycle Timeline
2DescriptionCVE.org
NVIDIA TensorRT for contains a vulnerability where an attacker might cause an improper validation of array index. A successful exploit of this vulnerability might lead to code execution.
Articles & Coverage 1
AnalysisAI
Improper array index validation (CWE-129) in NVIDIA TensorRT allows an attacker to trigger out-of-bounds memory access that may lead to arbitrary code execution when a victim processes malicious input on the local host. The CVSS 3.1 vector (AV:L/UI:R) indicates the target must actively load attacker-controlled content, so exploitation hinges on tricking a user or automated pipeline into ingesting a crafted model or input file. There is no public exploit identified at time of analysis and the CVE is not in CISA KEV, but with high confidentiality, integrity, and availability impact this is a meaningful priority for AI/ML inference environments.
Technical ContextAI
TensorRT is NVIDIA's high-performance deep-learning inference SDK and optimizer used to build and run optimized inference engines on NVIDIA GPUs, commonly consuming serialized network/model artifacts such as ONNX or serialized engine/plan files. CWE-129 (Improper Validation of Array Index) means the code uses an externally influenced value to index into an array or buffer without adequately bounding it, producing an out-of-bounds read or write. In a native C/C++ inference stack like TensorRT, such an out-of-bounds access during parsing or engine construction can corrupt memory and, per the advisory, escalate to code execution. The single affected CPE (cpe:2.3:a:nvidia:tensorrt:*) covers the TensorRT application/library across unspecified versions, so exact fixed builds must be taken from NVIDIA's advisory rather than the CPE wildcard.
RemediationAI
Upgrade TensorRT to the fixed release specified in NVIDIA's security bulletin at https://nvidia.custhelp.com/app/answers/detail/a_id/5855; the exact patched version is not present in the provided data and should be taken directly from that advisory rather than assumed (patch available per vendor advisory, released patched version not independently confirmed from the input). Until patched, treat every model, ONNX file, and serialized engine/plan as untrusted input: only load artifacts from sources you control or cryptographically verify, and disable or gate any automated pipelines that deserialize third-party models, accepting that this interrupts CI/CD or model-import automation. Run TensorRT workloads under least-privilege service accounts and in sandboxed or containerized inference environments so an out-of-bounds code-execution event is contained, at the cost of added operational complexity. Restrict who can submit models to shared inference or build hosts to shrink the UI:R trigger surface.
Local code execution in NVIDIA TensorRT is possible when the library parses an attacker-supplied input (such as a crafte
Code execution in NVIDIA TensorRT is possible when the SDK processes a maliciously crafted input that overflows a heap-b
Out-of-bounds write in NVIDIA TensorRT prior to version 10.16.1 allows an attacker to corrupt memory and tamper with dat
Code execution in NVIDIA TensorRT (all versions through v10.16.1) arises from unsafe deserialization of untrusted data (
Same weakness CWE-129 – Improper Validation of Array Index
View allShare
External POC / Exploit Code
Leaving vuln.today