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 2602.07107cs.CRcs.AI
收藏

ShallowJail: 对大语言模型的对抗性攻击

ShallowJail: Steering Jailbreaks against Large Language Models

xxx

Shang Liu, Hanyu Pei, Zeyan Liu

AI总结:

ShallowJail通过操纵LLMs的初始标记来攻击其对齐机制,从而降低大语言模型的安全性。

摘要:

大型语言模型(LLMs)在许多领域取得了成功。通常通过对齐来防止它们被用于有害目的。然而,对齐的LLMs仍然容易受到劫持攻击的影响,这些攻击故意误导它们生成有害的输出。现有的劫持攻击要么是黑盒攻击,使用精心设计但不隐蔽的提示,要么是白盒攻击,需要资源密集的计算。鉴于这些挑战,我们介绍了ShallowJail,一种新的攻击方法,利用LLMs中的浅层对齐。ShallowJail可以通过在推理过程中操纵初始标记来误导LLMs的响应。通过广泛的实验,我们展示了ShallowJail的有效性,证明其显著降低了最先进的LLMs响应的安全性。我们的代码可在https://github.com/liuup/ShallowJail上获得。

英文摘要:

Large Language Models(LLMs) have been successful in numerous fields. Alignment has usually been applied to prevent them from harmful purposes. However, aligned LLMs remain vulnerable to jailbreak attacks that deliberately mislead them into producing harmful outputs. Existing jailbreaks are either black-box, using carefully crafted, unstealthy prompts, or white-box, requiring resource-intensive computation. In light of these challenges, we introduce ShallowJail, a novel attack that exploits shallow alignment in LLMs. ShallowJail can misguide LLMs' responses by manipulating the initial tokens during inference. Through extensive experiments, we demonstrate the effectiveness of ShallowJail, which substantially degrades the safety of state-of-the-art LLM responses. Our code is available at https://github.com/liuup/ShallowJail.

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。

↑