PassGAN-Based Honeywords System

Abstract

In this work, we simulate a game between the attacker and the defender to evaluate their possible strategies in a honeyword system. For the attacker's guessing strategy, the Top-PW attack is used and a novel PassGAN-based attack is proposed. Meanwhile, for the defender's honeywords generation strategy, a PassGAN-based method is implemented and a novel two-stages PassGAN-based generation strategy is proposed. The experiment results show that the Top-PW attack is generally more successful than the PassGAN-based attack for the attacker while the proposed two-stages PassGAN-based approach is proven to be the most suitable honeyword generation strategy for the defender. The proposed two-stages PassGAN-based method for the defender's generation strategy can decrease the attacker's success rate to 9% and surpass some previous honeywords generation methods.