Explainable Autonomous Cyber Defense using Adversarial Multi-Agent Reinforcement Learning

📄 arXiv: 2604.04442 📥 PDF

作者: Yiyao Zhang, Diksha Goel, Hussain Ahmad

分类: cs.CR, cs.LG, cs.MA

发布日期: 2026-04-07

💡 一句话要点

提出Causal Multi-Agent Decision Framework以解决网络防御中的模糊性问题

🎯 匹配领域: 支柱二：RL算法与架构 (RL & Architecture)

关键词: 网络安全 自主代理 因果建模 强化学习 多代理系统 可解释性 误报率降低

📋 核心要点

  1. 现有的网络防御方法主要依赖于基于相关性的信号，缺乏结构约束，容易在模糊输入下出现推理漂移。
  2. C-MADF通过学习结构因果模型并将其编译为有向无环图，限制了响应动作的因果一致性，提升了决策的准确性。
  3. 在CICIoT2023数据集上，C-MADF将误报率从11.2%降低至1.8%，同时实现了0.997的精度和0.979的F1-score。

📝 摘要（中文）

随着自主代理在网络攻防中的广泛应用，复杂的闭环交互在关键基础设施环境中变得日益普遍。高级持续威胁（APT）行为者利用“依赖现有资源”的技术和针对性遥测扰动，导致监控系统产生模糊性，从而使自动防御系统过度反应或错误分类无害行为为恶意活动。现有的防御方法在应对模糊或对抗性输入时缺乏结构约束，容易出现推理漂移。为此，本文提出了Causal Multi-Agent Decision Framework（C-MADF），该框架结合了因果建模与对抗双策略控制，旨在提高网络防御的有效性和可解释性。通过在真实数据集上进行实验，C-MADF显著降低了误报率，并实现了高精度和高召回率。

🔬 方法详解

问题定义：本文旨在解决现有网络防御方法在面对模糊和对抗性输入时的不足，特别是推理漂移和误报率高的问题。

核心思路：C-MADF通过构建结构因果模型，限制决策空间中的响应动作，确保其因果一致性，从而提高网络防御的可靠性和可解释性。

技术框架：C-MADF的整体架构包括三个主要模块：首先，从历史遥测数据中学习结构因果模型；其次，编译成有向无环图（DAG）以定义可接受的响应转移；最后，利用双代理强化学习系统进行决策。

关键创新：C-MADF的主要创新在于结合了因果建模与对抗双策略控制，形成了一个结构约束的决策框架，显著提高了网络防御的有效性和可解释性。

关键设计：在设计中，采用了政策发散评分来量化代理间的不一致性，并通过人机交互界面提供可解释性透明度评分，作为不确定性下的升级信号。

🖼️ 关键图片

fig_0
fig_1
fig_2

📊 实验亮点

在CICIoT2023数据集上的实验结果显示，C-MADF将误报率从11.2%、9.7%和8.4%显著降低至1.8%。同时，该框架实现了0.997的精度、0.961的召回率和0.979的F1-score，展现了其在网络防御中的卓越性能。

🎯 应用场景

该研究在网络安全领域具有广泛的应用潜力，尤其是在关键基础设施的自动化防御系统中。通过提高防御系统的可解释性和准确性，C-MADF能够有效应对复杂的网络攻击，降低误报率，提升整体安全性。未来，该框架可扩展至其他领域，如智能制造和物联网安全等。

📄 摘要（原文）

Autonomous agents are increasingly deployed in both offensive and defensive cyber operations, creating high-speed, closed-loop interactions in critical infrastructure environments. Advanced Persistent Threat (APT) actors exploit "Living off the Land" techniques and targeted telemetry perturbations to induce ambiguity in monitoring systems, causing automated defenses to overreact or misclassify benign behavior as malicious activity. Existing monolithic and multi-agent defense pipelines largely operate on correlation-based signals, lack structural constraints on response actions, and are vulnerable to reasoning drift under ambiguous or adversarial inputs. We present the Causal Multi-Agent Decision Framework (C-MADF), a structurally constrained architecture for autonomous cyber defense that integrates causal modeling with adversarial dual-policy control. C-MADF first learns a Structural Causal Model (SCM) from historical telemetry and compiles it into an investigation-level Directed Acyclic Graph (DAG) that defines admissible response transitions. This roadmap is formalized as a Markov Decision Process (MDP) whose action space is explicitly restricted to causally consistent transitions. Decision-making within this constrained space is performed by a dual-agent reinforcement learning system in which a threat-optimizing Blue-Team policy is counterbalanced by a conservatively shaped Red-Team policy. Inter-policy disagreement is quantified through a Policy Divergence Score and exposed via a human-in-the-loop interface equipped with an Explainability-Transparency Score that serves as an escalation signal under uncertainty. On the real-world CICIoT2023 dataset, C-MADF reduces the false-positive rate from 11.2%, 9.7%, and 8.4% in three cutting-edge literature baselines to 1.8%, while achieving 0.997 precision, 0.961 recall, and 0.979 F1-score.