You need to agree to share your contact information to access this model

This repository is publicly accessible, but you have to accept the conditions to access its files and content.

Log in or Sign Up to review the conditions and access this model content.

TensorRT ONNX Parser Heap OOB Read PoC

Security research PoC β€” DO NOT USE AS A REAL MODEL.

Files

  • poc_heap_overflow.py β€” Python script that crafts malicious .onnx files and triggers OOB reads
  • poc_external_oob.onnx β€” Crafted ONNX model with external weights offset past file boundary
  • poc_double_oob.onnx β€” Crafted ONNX model with shape/data count mismatch (convertDouble OOB)
  • weights.bin β€” 16-byte external weights file used by poc_external_oob.onnx

Reproduce

pip install tensorrt==10.15.1 onnx
python poc_heap_overflow.py

What it does

Finding A: External weights with offset=999999 on a 16-byte file β€” parser succeeds silently with OOB pointer. Finding B: shape=[50,50] with only 4 doubles β€” convertDouble reads 2496 heap elements, garbage visible in TRT warnings.

Downloads last month

-

Downloads are not tracked for this model. How to track
Inference Providers NEW
This model isn't deployed by any Inference Provider. πŸ™‹ Ask for provider support