This project was born out of curiosity. I've been doing a lot of reading about AV, neural networks, and studying malware more in depth. Multiple papers from arxiv inspired this.
BinSleuth is an anti-malware scanner designed to hunt down malicious code in Windows binaries, ELF and Mach-O files, making it a versatile tool across platforms.
A little about the training. This model uses a FNN. It was trained on 20,000 files, both benign and malicous, achieving 94% accuracy rate during testing. But keep in mind, this is still in its alpha phase—so don't rely on it for production environments just yet, or ever! The training and inference code will not be released until later, but if you're interested please shoot me a dm on X and I'll be glad to provide it.(dev_null321)
Also, please don't complain about the lack of dynamic analysis, heuristics, etc. I know. That will be added later, but these things take time, and are even hard for big AV companies to implement properly. Enjoy!
Return to Home