My qualifying Master Thesis for Msc in Computer Science.
The thesis explores the impact of language differences on Generative Adversarial Networks when such systems are used as a password cracking tool. Specifically, we compare the cracking performance of a GAN model trained on English language passwords, with a model trained on Italian passwords and/or italian natural language samples. We find that GANs can be a valuable tool for tackling password cracking when working with data in another language, due to the GAN's ability to generate unbound numbers of password candidates. Ultimately they can be used effectively in conjunction wit traditional rule-based tools or in a standalone fashion.
A link to the paper can be found here.