Plexiglass: A toolbox for testing against adversarial attacks in DNNs and LLMs

Hi everyone, my name is Enoch and I am a researcher studying deep generative models.

I've started this project called Plexiglass a while back, which started off as a torch toolbox for adversarial research in DCNNs. I am now rebooting it as a toolbox for testing against adversarial attacks in both DNNs and LLMs.

Idea is to test your DCNNs against adversarial attacks such as fast gradient sign method and toxic prompts in LLMs.

I would very much appreciate contributions, I need more devs as I'm too busy to do this all by myself .

Repo is here: https://github.com/kortex-labs/plexiglass