PassGAN

Why we included this project

Security researchers and red-teamers doing password audits will find PassGAN a compact reference implementation of the Wasserstein GAN approach from the 2017 paper. The repo has a clean CLI for training on your own leaked password lists and for sampling candidates from a pretrained model trained on RockYou, so you can generate guesses without building the model yourself. Because it builds directly on the improved WGAN training implementation, it also works as a readable example of adapting adversarial training to a generative security task. Teams wondering whether generative models beat rule-based or Markov guessing tools can run it and compare coverage against their existing pipeline.