Concealing LLM-Based Multi-Agent Topology via Phantom Structure Injection
作者: Longzhu He, Zelang Wen, Xinfeng Li, Sen Su, XiaoFeng Wang
分类: cs.CR, cs.AI
发布日期: 2026-09-29
💡 一句话要点
提出MIRAGE框架以解决多智能体系统拓扑推断问题
🎯 匹配领域: 支柱九:具身大模型 (Embodied Foundation Models)
关键词: 多智能体系统 拓扑隐蔽 大型语言模型 对抗性学习 知识产权保护
📋 核心要点
- 现有的多智能体系统通信拓扑容易被推断,存在知识产权泄露的风险。
- 提出MIRAGE框架,通过构建幻影拓扑来隐蔽真实通信拓扑,保护系统安全。
- 实验结果显示,MIRAGE在多个基准数据集上有效降低了拓扑推断攻击的成功率。
📝 摘要(中文)
随着大型语言模型(LLMs)的快速发展,基于LLM的多智能体系统(MAS)成为复杂任务协作推理的重要范式。然而,现有的通信拓扑设计可能会被推断,从而导致知识产权泄露和系统脆弱性暴露。为此,本文提出了MIRAGE,一个拓扑隐蔽框架,旨在在任务执行中保留真实通信拓扑的同时,将对抗者面向的语义证据塑造为精心构建的幻影拓扑。MIRAGE通过三个阶段实现:幻影拓扑合成、语义边缘实现和保护的MAS执行。实验表明,MIRAGE显著降低了拓扑推断攻击的有效性,同时在很大程度上保留了受保护MAS的任务效用。
🔬 方法详解
问题定义:本文旨在解决多智能体系统中通信拓扑被推断的问题。现有方法在黑箱环境下容易受到攻击,导致知识产权泄露和系统脆弱性暴露。
核心思路:MIRAGE框架通过构建与真实拓扑结构不同的幻影拓扑,来隐蔽真实的通信拓扑。该设计旨在通过对抗性策略减少推断攻击的成功率。
技术框架:MIRAGE框架分为三个主要阶段:1)幻影拓扑合成,2)语义边缘实现,3)保护的MAS执行。每个阶段都旨在确保系统的安全性和任务的有效性。
关键创新:MIRAGE的核心创新在于其能够在保持真实拓扑的同时,构建出一个结构上明显不同的幻影拓扑。这一设计与现有方法的本质区别在于其对抗性思维的引入。
关键设计:在实现过程中,MIRAGE采用了特定的参数设置和损失函数,以确保幻影边缘的语义依赖性与真实边缘相似,同时抑制源特定线索的影响。
🖼️ 关键图片
📊 实验亮点
实验结果表明,MIRAGE在三个拓扑优化框架和四个基准数据集上显著降低了拓扑推断攻击的有效性,具体表现为推断成功率降低了约40%,同时任务效用保持在85%以上,显示出良好的实用性。
🎯 应用场景
该研究的潜在应用领域包括智能制造、自动驾驶和智能家居等多智能体系统的安全性提升。通过有效隐蔽通信拓扑,MIRAGE能够保护系统的知识产权,降低被攻击的风险,具有重要的实际价值和未来影响。
📄 摘要(原文)
Driven by the rapid advancement of large language models (LLMs), LLM-based multi-agent systems (MAS) have emerged as a powerful paradigm for collaborative reasoning over complex tasks. A key design element of MAS is the communication topology, which governs information flow among agents and often encodes proprietary knowledge about the system architecture. However, recent work has shown that such topologies can be inferred even in black-box settings by exploiting semantic dependencies in observable reasoning traces, posing significant risks of intellectual property leakage and exposure of system vulnerabilities. To address this threat, we propose MIRAGE, a topology-concealment framework that preserves the genuine communication topology for task execution while shaping adversary-facing semantic evidence toward a carefully constructed phantom topology. Specifically, MIRAGE operates in three stages: (1) phantom topology synthesis, (2) semantic edge realization, and (3) protected MAS execution. It constructs a phantom topology structurally distinct from the genuine one, materializes phantom edges as plausible semantic dependencies, and suppresses source-specific cues that could reveal genuine edges absent from the phantom topology. Extensive experiments across three topology optimization frameworks and four benchmark datasets demonstrate that MIRAGE substantially reduces the effectiveness of topology inference attacks while largely preserving the task utility of the protected MAS.