Unleashing Password Potential: VAEs vs GANs in a Password Guessing Duel

Insecure passwords remain a persistent vulnerability in cybersecurity. This study investigates the potential of employing Variational Autoencoders (VAEs) and Generative Adversarial Networks (GANs) as tools for generating synthetic passwords, with the aim of enhancing existing password cracking methodologies. These models, trained on a dataset of leaked RockYou passwords, are evaluated based on their capacity to produce passwords that replicate the complexity and diversity observed in real-world passwords. Through a thorough assessment of the quality and characteristics of these synthetic passwords, this research reveals the strengths and weaknesses of each model. Additionally, the study examines the effectiveness of utilizing synthetic passwords in password cracking endeavors, comparing their performance to established methods and actual breached passwords. In conclusion, this research seeks to deepen our comprehension of password creation patterns and fortify our ability to counter the risks associated with weak passwords. By delving into the potential of VAEs and GANs, we contribute to the ongoing pursuit of robust cybersecurity measures and the safeguarding of sensitive information.