CVE-2026-115922Medium· 6.2▾ SunlitONNX: Denial of service via unchecked expansion of packed sub-byte tensors in ONNX Python APIs
▾ Sunlit zone — Low / medium · no exploitation signal
impact 34.1 · likelihood 0 · exploitation 0
Need a working PoC? Pro members can cast a request and our team develops one — it lands right here.
Exploit-prediction probability, daily snapshots since Oct 27.
Disclosure to exploitation, from the record and what we observed since indexing it.
0.1%
Disclosed via OSV
ONNX's Python tensor-materialization helpers could allocate memory according to attacker-controlled tensor dimensions without verifying that a packed sub-byte payload contained the declared number of elements.
A very small model could therefore cause a disproportionately large allocation when an application called onnx.numpy_helper.to_array() or constructed onnx.reference.ReferenceEvaluator with the model. The issue affected UINT2, INT2, UINT4, INT4, and FLOAT4E2M1 tensors stored in either raw_data or packed int32_data.
The vulnerable _unpack_2bit() and _unpack_4bit() paths first allocated a small array based on the available packed bytes, then resized it to the shape declared in TensorProto.dims. They did not first confirm that the payload encoded enough logical elements for that shape.
For example, the reporter demonstrated a 66-byte model containing one packed byte and a declared dimension of 700,000,000. The model passed onnx.checker.check_model() and allocated approximately 700 MB while materializing the initializer, followed by another large allocation when executing the reference evaluator.
Loading, checking, or running shape inference on the serialized model alone did not trigger the allocation. The vulnerable tensor had to be materialized.
An application that materializes packed tensors from an untrusted model can exhaust available memory, causing denial of service. Memory corruption and code execution were not demonstrated.
Upgrade to ONNX 1.22.0 or later.
The primary fix validates the logical element count before resizing the unpacked NumPy array. A complementary checker change rejects an undersized packed payload before materialization.
The five affected tensor types and both storage formats were re-tested against the published ONNX 1.23.2 wheel and are rejected before the oversized allocation.
To verify the behavior using a released vulnerable package in an isolated environment, install onnx==1.21.0 with python -m pip install "onnx==1.21.0". Run memory-exhaustion demonstrations only with an appropriate process memory limit.
The vulnerable ONNX component processes the malicious model through a local library API or file boundary and is not itself bound to a network stack. Applications that expose ONNX model processing over a network should calculate an Environmental score with Modified Attack Vector set to Network.
onnx < 1.22.0Upgrade to a patched release:
onnx 1.22.0Connected by shared product, vendor, weakness, or advisory.
CVE-2026-115919Medium· 6.2ONNX: Out-of-Bounds Read in RNN/GRU/LSTM Shape Inference (Old Opsets, Missing Rank Check)
CVE-2026-115920Medium· 6.2ONNX: Out-of-bounds read in ConvTranspose shape inference via undersized weight rank
CVE-2026-115921Medium· 6.2ONNX: Null Pointer Dereference in ScanInferenceFunction Due to Missing Required Attribute
CVE-2026-44512Medium· 5.5ONNX has Null Pointer Dereference in Upsample Version Converter Adapter (Zero Inputs)
CVE-2024-27318High· 7.5Onnx Directory Traversal vulnerability
CVE-2026-34447Medium· 5.5ONNX: External Data Symlink Traversal