CVE-2026-55093

tract-nnef: integer overflow in NNEF `.dat` tensor parser yields an out-of-bounds read on model load

Published Sep 14, 2026 CVSS 3.1 · 6.1 Medium EPSS < 1% KEV · Not listed

Description

Tract is a tiny, no-nonsense, self-contained TensorFlow and ONNX inference toolkit. Prior to 0.21.16, 0.22.2, and 0.23.1, tract-nnef uses unchecked usize multiplication in nnef/src/tensors.rs read_tensor for attacker-controlled tensor dimensions, the allocation size, and the reported tensor length. Loading a crafted NNEF archive through model_for_path or model_for_read reaches the default DatLoader and can make the wrapped size check accept a small allocation while data/src/tensor.rs as_slice_unchecked creates a much larger logical slice. Model construction through as_uniform can then read beyond the heap allocation and disclose adjacent data, and later access can terminate the process with a segmentation fault. The affected dense numeric tensor path does not include the independently guarded bool, String, or block-quant paths, and no out-of-bounds write or code execution was demonstrated. This issue is fixed in versions 0.21.16, 0.22.2, and 0.23.1.

Analysis

Technical impact

Tract is a lightweight inference toolkit that parses NNEF of usize dimensions in the dense numeric tensor path; when a malicious NNEF archive supplies oversized dimensions, the multiplication overflows, causing an allocation that is smaller than the logical data slice. During loading via model_for_path or model_for_read, this leads to an out-of-bounds read that can expose adjacent memory and may trigger a segmentation fault. The bool, String, and block-quant tensor paths are independently guarded, and no out-of-bounds write or code execution was demonstrated.

Affected Systems

The flaw affects Sonos tract binaries in versions earlier than 0.21.16, 0.22.2, and 0.23.1. Any application that loads NNEF archives—either via model_for_path or model_for_read—without validating tensor dimensions is susceptible. Alternative tensor paths for bool, String, or block‑quant remain protected and are not impacted by this issue.

Risk and Exploitability

The CVSS score is 6.1, indicating moderate severity. The EPSS score is < 1 %, and the vulnerability is not listed in the CISA KEV catalog, suggesting a low to moderate exploitation probability once exposed. The attack requires the attacker to supply a malicious NNEF model file; therefore, the likely vector is local or within an environment that accepts untrusted model inputs. The impact includes sensitive data leakage from memory and potential service disruption through crashes, but not remote code execution.

Affected Products

Default status is the baseline for the product, each version can override it (e.g. patched versions marked unaffected).

Vendor Product Default status Versions

sonos

tract

affected

Version	Status	Constraints
`< 0.21.16`	affected	—
`>= 0.22.0, < 0.22.2`	affected	—
`>= 0.23.0, < 0.23.1`	affected	—

No data.

Vendor Product Versions

Sonos

Tract

Version	Status	Scheme	Platform
`[0,0.21.16)`	affected	semver	—
`[0.22.0,0.22.2)`	affected	semver	—
`[0.23.0,0.23.1)`	affected	semver	—

Remediation

No solution or workaround provided in the CVE record.

OpenCVE Recommended Actions

Upgrade tract to any of the patched versions (0.21.16, 0.22.2, or 0.23.1)
Restrict or validate the size and dimensions of input tensors before they are fed into tract, ensuring multiplication does not overflow
Run tract inside a sandboxed or limited‑privilege environment when processing untrusted models

Generated by OpenCVE AI on September 20, 2026 at 23:02 UTC.

Tracking

Advisories

Source	ID	Title
Github GHSA	GHSA-x5mv-8wgw-29hg	tract-nnef: integer overflow in NNEF `.dat` tensor parser yields an out-of-bounds read on model load

Metrics

No CVSS v4.0

Attack Vector Local

Attack Complexity Low

Privileges Required None

Scope Unchanged

Confidentiality Impact Low

Integrity Impact None

Availability Impact High

User Interaction Required

No CVSS v3.0

No CVSS v2

This CVE is not in the KEV list.

Exploit prediction score: < 1% (0.181% probability within 30 days)

Exploitation poc

Automatable no

Technical Impact partial

References

Link	Providers
https://github.com/sonos/tract/commit/34c7df2c9bd2a36583e09b52f3e6319bf23102e8
https://github.com/sonos/tract/commit/828ec20f74c14c0923c7d7f0cff062ed222d630a
https://github.com/sonos/tract/commit/ee5f019d39fb97f7d4431b883c3ce97b74c28a64
https://github.com/sonos/tract/releases/tag/0.21.16
https://github.com/sonos/tract/releases/tag/0.22.2
https://github.com/sonos/tract/releases/tag/v0.23.1
https://github.com/sonos/tract/security/advisories/GHSA-x5mv-8wgw-29hg

History

Sep 15, 2026 12:15 AM First appearance in the feed. Affected vendors updated View changes

Type	Values Removed	Values Added
First Time appeared		Sonos Sonos tract
Vendors & Products		Sonos Sonos tract

Sep 14, 2026 9:30 PM Scoring metrics updated View changes

Type	Values Removed	Values Added
Metrics		ssvc `{'options': {'Automatable': 'no', 'Exploitation': 'poc', 'Technical Impact': 'partial'}, 'version': '2.0.3'}`

Sep 14, 2026 9:00 PM Description updated. Title updated. Weaknesses updated. Added references. Scoring metrics updated View changes

Type	Values Removed	Values Added
Description		Tract is a tiny, no-nonsense, self-contained TensorFlow and ONNX inference toolkit. Prior to 0.21.16, 0.22.2, and 0.23.1, tract-nnef uses unchecked usize multiplication in nnef/src/tensors.rs read_tensor for attacker-controlled tensor dimensions, the allocation size, and the reported tensor length. Loading a crafted NNEF archive through model_for_path or model_for_read reaches the default DatLoader and can make the wrapped size check accept a small allocation while data/src/tensor.rs as_slice_unchecked creates a much larger logical slice. Model construction through as_uniform can then read beyond the heap allocation and disclose adjacent data, and later access can terminate the process with a segmentation fault. The affected dense numeric tensor path does not include the independently guarded bool, String, or block-quant paths, and no out-of-bounds write or code execution was demonstrated. This issue is fixed in versions 0.21.16, 0.22.2, and 0.23.1.
Title		tract-nnef: integer overflow in NNEF `.dat` tensor parser yields an out-of-bounds read on model load
Weaknesses		CWE-125 CWE-190
References		https://github.com/sonos/tract/commit/34c7df2c9bd2a36583e09b52f3e6319bf23102e8 https://github.com/sonos/tract/commit/828ec20f74c14c0923c7d7f0cff062ed222d630a https://github.com/sonos/tract/commit/ee5f019d39fb97f7d4431b883c3ce97b74c28a64 https://github.com/sonos/tract/releases/tag/0.21.16 https://github.com/sonos/tract/releases/tag/0.22.2 https://github.com/sonos/tract/releases/tag/v0.23.1 https://github.com/sonos/tract/security/advisories/GHSA-x5mv-8wgw-29hg
Metrics		cvssV3_1 `{'score': 6.1, 'vector': 'CVSS:3.1/AV:L/AC:L/PR:N/UI:R/S:U/C:L/I:N/A:H'}`