arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2310.02417cs.CR
收藏

狱中的越狱者:面向大语言模型的移动目标防御

Jailbreaker in Jail: Moving Target Defense for Large Language Models

Bocheng Chen, Advait Paliwal, Qiben Yan

AI总结:

针对大语言模型易受对抗攻击、难以兼顾有用性与无害性的问题,研究提出移动目标防御增强系统,通过多候选模型对齐输出与查询输出分析过滤,实现攻击成功率和拒绝率均降至0%。

摘要:

大语言模型(LLM)以理解和遵循指令的能力著称,但易受对抗攻击影响。研究人员发现,当前商用LLM在面对对抗性查询时,要么因给出不道德答案而做不到“无害”,要么因拒绝提供有意义答案而做不到“有用”。为在有用性与无害性之间取得平衡,我们设计了一套移动目标防御(MTD)增强的LLM系统。该系统旨在输出与多个候选模型输出对齐的无毒答案,使其对对抗攻击更具鲁棒性。我们设计了一个查询与输出分析模型,以过滤掉不安全或无响应的答案。我们使用当前最先进的对抗性查询,对8个最新的聊天机器人模型进行了评估。我们的MTD增强LLM系统将攻击成功率从37.5%降至0%,同时将响应拒绝率从50%降至0%。

英文摘要:

Large language models (LLMs), known for their capability in understanding and following instructions, are vulnerable to adversarial attacks. Researchers have found that current commercial LLMs either fail to be "harmless" by presenting unethical answers, or fail to be "helpful" by refusing to offer meaningful answers when faced with adversarial queries. To strike a balance between being helpful and harmless, we design a moving target defense (MTD) enhanced LLM system. The system aims to deliver non-toxic answers that align with outputs from multiple model candidates, making them more robust against adversarial attacks. We design a query and output analysis model to filter out unsafe or non-responsive answers. %to achieve the two objectives of randomly selecting outputs from different LLMs. We evaluate over 8 most recent chatbot models with state-of-the-art adversarial queries. Our MTD-enhanced LLM system reduces the attack success rate from 37.5\% to 0\%. Meanwhile, it decreases the response refusal rate from 50\% to 0\%.

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。

↑