基于多智能体的网络入侵检测系统的研究  

Research on Multi-Agent Based Network Intrusion Detection Systems

作　　者：张利庭 周恩祥 吴静妹 ZHANG Liting;ZHOU Enxiang;WU Jingmei(School of Computer Science and Artificial Intelligence,Wanjiang University of Technology,Ma'anshan Anhui 243031,China)

机构地区：[1]皖江工学院计算机与人工智能学院,安徽马鞍山243031

出　　处：《佳木斯大学学报(自然科学版)》2024年第12期44-47,共4页Journal of Jiamusi University:Natural Science Edition

基　　金：2024年度安徽省高校自然科学研究重点项目(2024AH051871);2023年度安徽省高校自然科学研究重点项目(2023AH052492);2022年安徽省高等学校省级质量工程校企合作项目(2022xqhz085)。

摘　　要：针对传统网络入侵检测系统在识别复杂多变的网络攻击模式时准确率低的问题,提出了一种基于独立Q学习的多智能体网络入侵检测系统。该系统采用多智能体架构,每个智能体运用独立Q学习算法对网络行为进行监控和风险评估,通过智能化的方法提升网络入侵检测的效率和准确性。实验结果表明,与传统入侵检测方法相比,本系统的检测准确率、响应速度和鲁棒性等性能均有显著提升,能够有效识别未知攻击模式,展现出良好的自适应性和泛化能力,为网络安全领域提供了一种新的解决方案,对于构建更加智能和可靠的网络安全防护体系具有一定理论和实践价值。Aiming at the problem of low accuracy of traditional network intrusion detection system in recognizing complex and variable network attack patterns,a multi-intelligence body network intrusion detection system based on independent Q-learning is proposed.The system adopts multi-intelligent body architecture,and each intelligent body applies independent Q-learning algorithms to monitor network behaviors and risk assessment,which improves the efficiency and accuracy of network intrusion detection through an intelligent approach.The experimental results show that compared with the traditional intrusion detection methods,the detection accuracy,response speed and robustness of this system have been significantly improved,and it can effectively identify unknown attack patterns,showing good adaptive and generalization capabilities,providing a new solution for the field of network security,and it has a certain theoretical and practical value for the construction of a more intelligent and reliable network security protection system.

关 键 词：网络安全 入侵检测 强化学习 独立Q学习 

分 类 号：TP393.08[自动化与计算机技术—计算机应用技术]