arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2510.03417cs.CRcs.AI
收藏

NEXUS:在多轮 LLM 越狱中利用不安全序列的网络探索

NEXUS: Network Exploration for eXploiting Unsafe Sequences in Multi-Turn LLM Jailbreaks

Old Dominion University(旧 Dominion 大学);University of Arizona(亚利桑那大学)

Javad Rafiei Asl, Sidhant Narula, Mohammad Ghasemigol, Eduardo Blanco, Daniel Takabi

AI总结:

针对多轮 LLM 越狱中对抗查询探索不足的问题,本文提出 NEXUS,通过语义网络生成、反馈式查询链优化和自适应网络遍历提升攻击成功率。

摘要:

大语言模型(LLM)已经革新了自然语言处理,但仍易受越狱攻击影响,尤其是多轮越狱:这类攻击将恶意意图分散到看似良性的交流中,从而绕过对齐机制。现有方法通常对对抗空间的探索不足,依赖手工启发式规则,或缺乏系统化的查询优化。我们提出 NEXUS(Network Exploration for eXploiting Unsafe Sequences),这是一个用于构建、优化并执行多轮攻击的模块化框架。NEXUS 包括:(1)ThoughtNet,它将有害意图分层扩展为由主题、实体和查询链组成的结构化语义网络;(2)反馈驱动的 Simulator,通过攻击者—受害者—裁判 LLM 协作,并利用有害性和语义相似度基准,迭代式地优化和剪枝这些查询链;(3)Network Traverser,它在实时攻击中自适应地穿越经过优化的查询空间。该流程能够在不同 LLM 上发现隐蔽且高成功率的对抗路径。在若干闭源和开源 LLM 上,NEXUS 相比以往方法将攻击成功率提高了 2.1% 至 19.4%。代码:https://github.com/inspire-lab/NEXUS

英文摘要:

Large Language Models (LLMs) have revolutionized natural language processing but remain vulnerable to jailbreak attacks, especially multi-turn jailbreaks that distribute malicious intent across benign exchanges and bypass alignment mechanisms. Existing approaches often explore the adversarial space poorly, rely on hand-crafted heuristics, or lack systematic query refinement. We present NEXUS (Network Exploration for eXploiting Unsafe Sequences), a modular framework for constructing, refining, and executing optimized multi-turn attacks. NEXUS comprises: (1) ThoughtNet, which hierarchically expands a harmful intent into a structured semantic network of topics, entities, and query chains; (2) a feedback-driven Simulator that iteratively refines and prunes these chains through attacker-victim-judge LLM collaboration using harmfulness and semantic-similarity benchmarks; and (3) a Network Traverser that adaptively navigates the refined query space for real-time attacks. This pipeline uncovers stealthy, high-success adversarial paths across LLMs. On several closed-source and open-source LLMs, NEXUS increases attack success rate by 2.1% to 19.4% over prior methods. Code: https://github.com/inspire-lab/NEXUS

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。

↑