BinProv

BinProv is a tool for identifying compilation provenance by determining which compiler and optimization level was used to compile a binary, all without disassembly. It employs a BERT-style byte-level encoding of the raw bytes from the binary's `.text` section into a classification model, effectively covering architectures like x86_64, x86_32, ARM, and MIPS. Notable features include a pretrained model for various optimization levels, high accuracy on classification tasks, and detailed documentation for reproduction and inference.