arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2501.07959cs.AI
收藏

自指令少样本越狱:将攻击分解为模式与行为学习

Self-Instruct Few-Shot Jailbreaking: Decompose the Attack into Pattern and Behavior Learning

Technical University of Munich(慕尼黑工业大学);Shanghai Jiao Tong University(上海交通大学)

Jiaqi Hua, Wanxu Wei

AI总结:

针对少样本越狱(FSJ)在越狱高级模型时上下文过长的问题,本文提出配备演示级贪婪搜索的自指令少样本越狱(Self-Instruct-FSJ)框架,将攻击分解为模式与行为学习,更泛化高效地利用模型漏洞。

摘要:

最近,有几项研究致力于利用少样本恶意演示对大型语言模型(LLMs)进行越狱。特别是,Zheng等人专注于通过向演示中注入特殊标记并采用演示级随机搜索来提高少样本越狱(FSJ)的效率,即改进的少样本越狱(I-FSJ)。然而,我们注意到该方法在越狱高级模型时可能仍需要较长的上下文,例如对Meta-Llama-3-8B-Instruct(Llama-3)需要32个样本演示。在本文中,我们讨论了I-FSJ的局限性,并提出了配备演示级贪婪搜索的自指令少样本越狱(Self-Instruct-FSJ)。该框架将FSJ攻击分解为模式学习和行为学习,以更泛化和高效的方式利用模型的漏洞。我们进行了详尽的实验,在常见的开源模型上评估了我们的方法,并将其与基线算法进行比较。代码可在https://github.com/iphosi/Self-Instruct-FSJ获取。

英文摘要:

Recently, several works have been conducted on jailbreaking Large Language Models (LLMs) with few-shot malicious demos. In particular, Zheng et al. focus on improving the efficiency of Few-Shot Jailbreaking (FSJ) by injecting special tokens into the demos and employing demo-level random search, known as Improved Few-Shot Jailbreaking (I-FSJ). Nevertheless, we notice that this method may still require a long context to jailbreak advanced models e.g. 32 shots of demos for Meta-Llama-3-8B-Instruct (Llama-3) \cite{llama3modelcard}. In this paper, we discuss the limitations of I-FSJ and propose Self-Instruct Few-Shot Jailbreaking (Self-Instruct-FSJ) facilitated with the demo-level greedy search. This framework decomposes the FSJ attack into pattern and behavior learning to exploit the model's vulnerabilities in a more generalized and efficient way. We conduct elaborate experiments to evaluate our method on common open-source models and compare it with baseline algorithms. Our code is available at https://github.com/iphosi/Self-Instruct-FSJ.

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。

↑