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 2502.00735cs.CRcs.AIcs.SE
收藏

从提示到扰动:针对音频大语言模型的自适应语音越狱框架

`From Prompt to Perturbation': An Adaptive Framework for Voice-Based Jailbreaks on Audio LLMs

School of Electrical and Computer Engineering, The University of Sydney(悉尼大学电气与计算机工程学院);University of Chicago(芝加哥大学);University of Texas at San Antonio(德克萨斯大学圣安东尼奥分校)

Linghan Huang, Bo Li, Huaming Chen, Kim-Kwang Raymond Choo

AI总结:

针对现有音频越狱攻击研究仅聚焦单一范式的不足,提出自适应越狱框架,可在统一设置下评估级联流水线和LALM,实验显示其攻击成功率优于现有方法。

摘要:

随着大语言模型(LLM)越来越多地集成到基于音频的应用中,人们对其抵御音频对抗攻击的脆弱性的担忧日益加剧。这些系统通常遵循两种架构范式:级联流水线(自动语音识别将音频输入转换为文本后再由LLM处理)和端到端的大型音频语言模型(LALM,直接解释原始音频信号)。除架构差异外,级联流水线主要易受通过语音传递的文本级越狱策略的攻击,而端到端LALM则引入了额外的声学-语义攻击向量。然而,现有研究往往仅聚焦于单一范式,对更广泛的音频攻击空间覆盖有限。为填补这一空白,我们提出一种自适应越狱攻击框架,用于在统一实验设置下系统评估级联流水线和LALM。该框架的核心是一个反馈引导的变异引擎,可自动生成并优化涵盖文本提示和音频扰动的越狱候选方案,从而扩大攻击的多样性和覆盖范围。对6个代表性音频系统的实验表明,两种范式均对音频越狱攻击存在显著脆弱性。与现有最优方法相比,我们的框架在各类音频大语言模型系统上均实现了更高的攻击成功率。

英文摘要:

As large language models (LLMs) are increasingly integrated into audio-based applications, growing concerns have emerged regarding their vulnerability to audio-based adversarial attacks. These systems typically follow two architectural paradigms: cascaded pipelines, where automatic speech recognition converts audio inputs into text before LLM processing, and end-to-end large audio-language models (LALMs), which directly interpret raw audio signals. Beyond architectural differences, cascaded pipelines are primarily vulnerable to text-level jailbreak strategies delivered through speech, whereas end-to-end LALMs introduce additional acoustic-semantic attack vectors. However, existing studies often focus on a single paradigm and provide limited coverage of the broader audio attack space. To bridge this gap, we propose an adaptive jailbreak attack framework for systematic evaluation of both cascaded pipelines and LALMs under a unified experimental setting. At its core, the framework uses a feedback-guided mutation engine to automatically generate and refine jailbreak candidates across both textual prompts and audio perturbations, thereby expanding attack diversity and coverage. Experiments on six representative audio-based systems demonstrate that both paradigms remain substantially vulnerable to audio jailbreak attacks. Compared with state-of-the-art methods, our framework achieves consistently higher attack success rates across diverse audio-based LLM systems.

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。

↑