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 2402.11753cs.CLcs.AI
收藏

ArtPrompt:基于ASCII艺术的针对对齐大语言模型的越狱攻击

ArtPrompt: ASCII Art-based Jailbreak Attacks against Aligned LLMs

University of Washington(华盛顿大学);Western Washington University(西华盛顿大学);University of Chicago(芝加哥大学)

Fengqing Jiang, Zhangchen Xu, Luyao Niu, Zhen Xiang, Bhaskar Ramasubramanian, Bo Li, Radha Poovendran

AI总结:

本文提出ArtPrompt越狱攻击,利用LLM难以识别ASCII艺术提示的弱点绕过安全对齐,并在五个主流LLM上验证其有效性。

摘要:

安全对于大语言模型(LLMs)的使用至关重要。人们已经开发了数据过滤和监督微调等多种技术来增强LLM的安全性。然而,目前已知的技术都假定用于LLM安全对齐的语料仅通过语义进行解释。但这一假设在现实世界应用中并不成立,这导致LLM存在严重漏洞。例如,论坛用户经常使用ASCII艺术(一种基于文本的艺术形式)来传递图像信息。在本文中,我们提出了一种新颖的基于ASCII艺术的越狱攻击,并引入了一个综合基准测试Vision-in-Text Challenge(ViTC),用于评估LLM识别无法仅通过语义解释的提示的能力。我们表明,五个最先进的LLM(GPT-3.5、GPT-4、Gemini、Claude和Llama2)难以识别以ASCII艺术形式提供的提示。基于这一观察,我们开发了越狱攻击ArtPrompt,它利用LLM在识别ASCII艺术方面的较差表现来绕过安全措施并诱导LLM产生不良行为。ArtPrompt仅需要对受害LLM进行黑盒访问,使其成为一种实用的攻击。我们在五个最先进的LLM上评估了ArtPrompt,并表明ArtPrompt能够有效且高效地诱导所有五个LLM产生不良行为。我们的代码可在https://github.com/uw-nsl/ArtPrompt获取。

英文摘要:

Safety is critical to the usage of large language models (LLMs). Multiple techniques such as data filtering and supervised fine-tuning have been developed to strengthen LLM safety. However, currently known techniques presume that corpora used for safety alignment of LLMs are solely interpreted by semantics. This assumption, however, does not hold in real-world applications, which leads to severe vulnerabilities in LLMs. For example, users of forums often use ASCII art, a form of text-based art, to convey image information. In this paper, we propose a novel ASCII art-based jailbreak attack and introduce a comprehensive benchmark Vision-in-Text Challenge (ViTC) to evaluate the capabilities of LLMs in recognizing prompts that cannot be solely interpreted by semantics. We show that five SOTA LLMs (GPT-3.5, GPT-4, Gemini, Claude, and Llama2) struggle to recognize prompts provided in the form of ASCII art. Based on this observation, we develop the jailbreak attack ArtPrompt, which leverages the poor performance of LLMs in recognizing ASCII art to bypass safety measures and elicit undesired behaviors from LLMs. ArtPrompt only requires black-box access to the victim LLMs, making it a practical attack. We evaluate ArtPrompt on five SOTA LLMs, and show that ArtPrompt can effectively and efficiently induce undesired behaviors from all five LLMs. Our code is available at https://github.com/uw-nsl/ArtPrompt.

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。

↑