This repository contains a Proof-of-Concept (PoC) model file (gguf_overflow.gguf) demonstrating a critical Integer Overflow vulnerability in standard GGUF parsers (like llama.cpp/ggml).
The GGUF format allows defining tensor dimensions. If a tensor is defined with dimensions that fit within a 64-bit element count but result in a byte size exceeding $2^{64}$ when multiplied by the data type size (e.g., Float32… See the full description on the dataset page:
https://huggingface.co/datasets/CandyCaine/GGUF-Integer-Overflow-PoC.
