ONNX

Products

ProductCVEsExploitedLatest
ONNX14–Oct 26, 2026
onnx/onnx2–Mar 20, 2025
onnx-mlir1–Jun 5, 2026

Exploited and recent CVEs

  • CVE-2026-115922 Medium 6.2 ONNX: Denial of service via unchecked expansion of packed sub-byte tensors in ONNX Python APIs 0.1%
  • CVE-2026-115921 Medium 6.2 ONNX: Null Pointer Dereference in ScanInferenceFunction Due to Missing Required Attribute 0.1%
  • CVE-2026-115920 Medium 6.2 ONNX: Out-of-bounds read in ConvTranspose shape inference via undersized weight rank 0.2%
  • CVE-2026-115919 Medium 6.2 ONNX: Out-of-Bounds Read in RNN/GRU/LSTM Shape Inference (Old Opsets, Missing Rank Check) 0.2%
  • CVE-2026-49114 High 7.1 ONNX symlink-following and path-traversal arbitrary file write 0.2%
  • CVE-2026-63632 Low 3.3 ONNX: Heap-Buffer-Overflow READ in Gemm Version Converter Adapter via Undersized Input Shape 0.2%
  • CVE-2026-44512 Medium 5.5 ONNX: Null Pointer Dereference in Upsample Version Converter Adapter (Zero Inputs) 0.2%
  • CVE-2026-11329 Low 3.6 onnx onnx-mlir Placeholder Node Cache backend.py generate_hash_key weak hash <0.1%
  • CVE-2026-34447 Medium 5.5 ONNX: External Data Symlink Traversal 0.3%
  • CVE-2026-34446 Medium 4.7 ONNX: Arbitrary File Read via ExternalData Hardlink Bypass in ONNX load 0.2%
  • CVE-2026-27489 High 8.6 ONNX: Path Traversal via Symlink 0.6%
  • CVE-2026-34445 High 8.6 ONNX: Malicious ONNX models can crash servers by exploiting unprotected object settings. 0.3%
  • CVE-2026-28500 High 8.6 ONNX Untrusted Model Repository Warnings Suppressed by silent=True in onnx.hub.load() — Silent Supply-Chain Attack 0.3%
  • CVE-2024-7776 High 8.1 Arbitrary File Overwrite in onnx/onnx 1.5%
  • CVE-2024-5187 High 8.8 Arbitrary File Overwrite in download_model_with_test_data in onnx/onnx 1.2%
  • CVE-2024-27319 Medium 4.4 Versions of the package onnx before and including 1.15.0 are vulnerable to Out-of-bounds Read as the ONNX_ASSERT and ONNX_ASSERTM functions have an off by one string copy. 0.6%
  • CVE-2024-27318 High 7.5 Versions of the package onnx before and including 1.15.0 are vulnerable to Directory Traversal as the external_data field of the tensor proto can have a path to the file which is outside the model cur 1.2%