arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2501.01830cs.CRcs.AIcs.CL
收藏

Auto-RT:面向大语言模型红队测试的自动越狱策略探索

Auto-RT: Automatic Jailbreak Strategy Exploration for Red-Teaming Large Language Models

Institute of Software, Chinese Academy of Sciences(中国科学院软件研究所);University of Chinese Academy of Sciences(中国科学院大学);Ant Group(蚂蚁集团)

Yanjiang Liu, Shuhen Zhou, Yaojie Lu, Huijia Zhu, Weiqiang Wang, Hongyu Lin, Ben He, Xianpei Han, Le Sun

AI总结:

Auto-RT 提出基于强化学习的自动越狱策略探索框架,通过提前终止探索和渐进奖励追踪机制高效发现大语言模型更广泛的安全漏洞,检测速度更快且成功率提升16.63%。

摘要:

自动化红队测试已成为发现大语言模型(LLMs)漏洞的关键方法。然而,现有大多数方法只关注孤立的安全缺陷,限制了它们适应动态防御和高效发现复杂漏洞的能力。为应对这一挑战,我们提出 Auto-RT,一个强化学习框架,能够自动探索并优化复杂攻击策略,通过恶意查询有效揭示安全漏洞。具体来说,我们引入两个关键机制以降低探索复杂度并改进策略优化:1)提前终止探索(Early-terminated Exploration),通过聚焦高潜力攻击策略来加速探索;2)带有中间降级模型的渐进奖励追踪算法(Progressive Reward Tracking),动态细化搜索轨迹以成功利用漏洞。在多种大语言模型上的大量实验表明,通过显著提升探索效率并自动优化攻击策略,Auto-RT 能检测到更广泛的漏洞,实现更快的检测速度,并且与现有方法相比成功率提高了16.63%。

英文摘要:

Automated red-teaming has become a crucial approach for uncovering vulnerabilities in large language models (LLMs). However, most existing methods focus on isolated safety flaws, limiting their ability to adapt to dynamic defenses and uncover complex vulnerabilities efficiently. To address this challenge, we propose Auto-RT, a reinforcement learning framework that automatically explores and optimizes complex attack strategies to effectively uncover security vulnerabilities through malicious queries. Specifically, we introduce two key mechanisms to reduce exploration complexity and improve strategy optimization: 1) Early-terminated Exploration, which accelerate exploration by focusing on high-potential attack strategies; and 2) Progressive Reward Tracking algorithm with intermediate downgrade models, which dynamically refine the search trajectory toward successful vulnerability exploitation. Extensive experiments across diverse LLMs demonstrate that, by significantly improving exploration efficiency and automatically optimizing attack strategies, Auto-RT detects a boarder range of vulnerabilities, achieving a faster detection speed and 16.63\% higher success rates compared to existing methods.

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。

↑