arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2306.01545cs.CLcs.AIcs.CR
收藏

PassGPT:基于大语言模型的密码建模与(引导式)生成

PassGPT: Password Modeling and (Guided) Generation with Large Language Models

ETH Zürich(苏黎世联邦理工学院);Swiss Data Science Center(瑞士数据科学中心);SRI International(SRI国际)

Javier Rando, Fernando Perez-Cruz, Briland Hitaj

AI总结:

该研究提出基于密码泄露数据训练的大语言模型PassGPT,其密码猜测能力是现有GAN方法的两倍,还支持引导式生成符合任意约束的密码,并可用于增强密码强度估计。

摘要:

大语言模型(LLM)无需显式监督,就能从海量文本中成功建模自然语言。本文研究了LLM在密码建模中的有效性。我们提出PassGPT——一种在密码泄露数据上训练的、用于密码生成的LLM。PassGPT的表现优于现有基于生成对抗网络(GAN)的方法,能猜出的未见过的密码数量是后者的两倍。此外,我们提出引导式密码生成的概念,利用PassGPT的采样流程生成符合任意约束的密码,这是当前基于GAN的策略不具备的能力。最后,我们深入分析了PassGPT在密码上定义的熵和概率分布,并探讨了它们在增强现有密码强度估计器方面的用途。

英文摘要:

Large language models (LLMs) successfully model natural language from vast amounts of text without the need for explicit supervision. In this paper, we investigate the efficacy of LLMs in modeling passwords. We present PassGPT, a LLM trained on password leaks for password generation. PassGPT outperforms existing methods based on generative adversarial networks (GAN) by guessing twice as many previously unseen passwords. Furthermore, we introduce the concept of guided password generation, where we leverage PassGPT sampling procedure to generate passwords matching arbitrary constraints, a feat lacking in current GAN-based strategies. Lastly, we conduct an in-depth analysis of the entropy and probability distribution that PassGPT defines over passwords and discuss their use in enhancing existing password strength estimators.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑