cve.bz

CVE-2026-63632 onnx onnx: Out-of-bounds Read

3.3 lowEPSS 0.2%Fix available

Low (3.3) in onnx onnx. EPSS 0.2% estimated probability of exploitation activity in the next 30 days. Fixed in onnx 1.22.0.

On this page
Open Neural Network Exchange (ONNX) is an open standard for machine learning interoperability. From 1.3.0 until 1.22.0, onnx.version_converter.convert_version() can perform an out-of-bounds read in Gemm_7_6::adapt_gemm_7_6() in onnx/version_converter/adapters/gemm_7_6.h when a Gemm node has input tensors with fewer than two dimensions because B_shape[1], A_shape[0], or A_shape[1] is accessed without a rank check, potentially causing a process crash during an opset 7 to 6 downgrade. This issue is fixed in version 1.22.0.

Severity

CVSS 3.1

3.3 low
CVSS:3.1/AV:L/AC:L/PR:N/UI:R/S:U/C:N/I:N/A:L
EPSS
0.2% (unchanged from 0.2% a week ago; 0.2% 30 days ago) estimated probability of exploitation activity in the next 30 days; 6th percentile. Scored 2026-11-22. A forecast for this CVE across the internet, not for your own systems.
CISA's assessment (SSVC)
Exploitation: PoC · Automatable: no · Technical impact: partial, assessed 2026-08-18. From CISA's Vulnrichment in the CVE List: whether it is exploited, whether exploiting it can be automated, and how much control it gives.
CWE
CWE-125 Out-of-bounds Read
Published
Modified
NVD analysis
Awaiting analysis
Record from
the NVD, synced

Fix and patched versions

Ask an AI about this CVE

Show the promptExplain CVE-2026-63632 for an engineer deciding whether to act. Use https://cve.bz/api/public/cves/CVE-2026-63632?via=ask-ai-copy and https://cve.bz/cve/CVE-2026-63632?via=ask-ai-copy as the source of truth: what is affected and which versions fix it, whether it is in CISA KEV, public exploits, and its EPSS (an internet-wide 30-day estimate of exploitation activity, not a forecast for my systems). If they give no fixed versions, use the CVE record https://cveawg.mitre.org/api/cve/CVE-2026-63632 or the vendor advisory, not third-party mirrors. Tell me how to check whether I am affected and how to verify a fix. Cite the sources and say where they are uncertain.

Opens a third-party assistant; its answer is its own, not cve.bz's. The prompt holds only this CVE's id and public links.

Affected

Versions

From the CNA's record in the CVE List.

ProductAffected
onnx>= 1.3.0, < 1.22.0

Packages (OSV)

AdvisoryEcosystemPackageAffectedFixed in
GHSA-p893-rvq9-2xf9PyPIonnx>=1.3.0 <1.22.01.22.0
PYSEC-2026-3587PyPIonnx>=1.3.0 <1.22.01.22.0

Linux distributions

DistributionReleasePackageFixed inAdvisory
Debian12, 13, 14onnxno fix listedDEBIAN-CVE-2026-63632
Ubuntu22.04:LTS, 26.04:LTS, 26.10 and 1 moreonnxno fix listedUBUNTU-CVE-2026-63632
Azure Linux3pytorch2.2.2-17AZL-96728

From the distributions' data via OSV.dev: the Debian Security Tracker; Ubuntu Security Notices © Canonical Ltd., CC BY-SA 4.0; Microsoft's Azure Linux vulnerability data, MIT. Modified: reformatted and linked to CVE records. Credits. On the distributions' own pages: Debian · Ubuntu.

Red Hat

Red Hat rates this Low.

Red Hat's statement

A flaw was found in the Open Neural Network Exchange (ONNX) library, which is bundled in several Red Hat OpenShift AI container images. The onnx.version_converter.convert_version() function performs an out-of-bounds read in the Gemm_7_6::adapt_gemm_7_6() adapter in onnx/version_converter/adapters/gemm_7_6.h when converting a model that contains a Gemm node whose input tensors have fewer than two dimensions, because B_shape[1], A_shape[0], or A_shape[1] is accessed without a rank check. An attacker who supplies a specially crafted ONNX model to an application that downgrades it from opset 7 to opset 6 can cause the process to crash, resulting in a denial of service. This is rated low severity (CVSS 3.3, AV:L/AC:L/PR:N/UI:R/S:U/C:N/I:N/A:L): exploitation requires a local user to process an untrusted model and only availability of the converting process is impacted; confidentiality and integrity are not affected. Of the Red Hat OpenShift AI components evaluated for this flaw, those that bundle a vulnerable onnx release are affected, while the remaining components are not affected because the vulnerable code is not present in them. Because this is a low-severity, availability-only issue, updates to the affected Red Hat OpenShift AI components are deferred; the residual risk is limited to a crash of the process that converts an untrusted model, which upstream has already fixed in its project releases.

Red Hat's mitigation

Mitigation for this issue is either not available or the currently available options do not meet the Red Hat Product Security criteria comprising ease of use and deployment, applicability to widespread installation base, or stability.

ProductPackageStateAdvisory
Red Hat OpenShift AI (RHOAI)rhoai/odh-mlserver-cuda-rhel9Not affected
Red Hat OpenShift AI (RHOAI)rhoai/odh-mlserver-rhel9Fix deferred
Red Hat OpenShift AI (RHOAI)rhoai/odh-openvino-model-server-rhel9Fix deferred
Red Hat OpenShift AI (RHOAI)rhoai/odh-pipeline-runtime-datascience-cpu-py312-rhel9Fix deferred
Red Hat OpenShift AI (RHOAI)rhoai/odh-pipeline-runtime-pytorch-cuda-py312-rhel9Fix deferred
Red Hat OpenShift AI (RHOAI)rhoai/odh-pipeline-runtime-pytorch-llmcompressor-cuda-py312-rhel9Fix deferred
Red Hat OpenShift AI (RHOAI)rhoai/odh-pipeline-runtime-pytorch-rocm-py312-rhel9Fix deferred
Red Hat OpenShift AI (RHOAI)rhoai/odh-pipeline-runtime-tensorflow-cuda-py312-rhel9Fix deferred
Red Hat OpenShift AI (RHOAI)rhoai/odh-pipeline-runtime-tensorflow-rocm-py312-rhel9Fix deferred
Red Hat OpenShift AI (RHOAI)rhoai/odh-workbench-codeserver-datascience-cpu-py312-rhel9Fix deferred
Red Hat OpenShift AI (RHOAI)rhoai/odh-workbench-jupyter-datascience-cpu-py312-rhel9Fix deferred
Red Hat OpenShift AI (RHOAI)rhoai/odh-workbench-jupyter-pytorch-cuda-py312-rhel9Fix deferred
Red Hat OpenShift AI (RHOAI)rhoai/odh-workbench-jupyter-pytorch-llmcompressor-cuda-py312-rhel9Fix deferred
Red Hat OpenShift AI (RHOAI)rhoai/odh-workbench-jupyter-pytorch-rocm-py312-rhel9Fix deferred
Red Hat OpenShift AI (RHOAI)rhoai/odh-workbench-jupyter-tensorflow-cuda-py312-rhel9Fix deferred
Red Hat OpenShift AI (RHOAI)rhoai/odh-workbench-jupyter-tensorflow-rocm-py312-rhel9Fix deferred
Red Hat OpenShift AI (RHOAI)rhoai/odh-workbench-jupyter-trustyai-cpu-py312-rhel9Fix deferred

Contains Red Hat Security Data © Red Hat, Inc., licensed under CC BY 4.0. Source: https://access.redhat.com/security/data. Modified: reformatted. Red Hat's page for CVE-2026-63632.

Attack patterns (CAPEC)

  • CAPEC-540 Overread Buffers Likelihood: Low · Severity: High

The attack patterns MITRE's CAPEC relates to this record's weaknesses, as MITRE names them. Credits.

References

Check your stackJSON