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.

YAML Metadata Warning:empty or missing yaml metadata in repo card

Check out the documentation for more information.

TensorRT RPROIPlugin Integer Overflow – PoC

Vulnerability: Integer overflow in RPROIPlugin deserialization → GPU heap overflow
File: plugin/nvFasterRCNN/nvFasterRCNNPlugin.cpp:102
Affects: TensorRT OSS ≤ 10.6 (NVIDIA)
Format: .engine / .trt

Files

File	Description
`malicious_rproi.bin`	Crafted plugin serialization blob (512 KB). Contains RPROIParams with `anchorsRatioCount = anchorsScaleCount = 65536` triggering integer overflow to 0, then 68 GB GPU write into 0-byte allocation.
`poc.py`	Reproducer — loads the blob directly via TRT plugin API

Reproduction

pip install tensorrt
python3 poc.py

Expected result: GPU heap overflow → CUDA error / process crash during deserializePlugin.

How it works

4 * 65536 * 65536 computed in int32_t overflows to 0.
cudaMalloc(0) succeeds with a minimal GPU allocation.
generateAnchors() then writes 68 GB into that allocation.

4 × anchorsRatioCount × anchorsScaleCount × sizeof(float)
= 4 × 65536 × 65536 × 4
= 68,719,476,736 bytes  →  truncated to int32  →  0
cudaMalloc(0) → ~512 byte GPU alloc
generateAnchors writes 68 GB → GPU heap overflow

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. 🙋 Ask for provider support