基于SOM网络的智能入侵检测系统  

An Intelligent Network Intrusion Detection System Based on SOM

作　　者：许素霞[1] 宋晓宇[1] 傅秀芬[1] 陈长瑶[1] 

机构地区：[1]广东工业大学计算机学院,广东广州510006

出　　处：《计算机仿真》2008年第9期111-114,共4页Computer Simulation

基　　金：广东省自然科学基金项目(06021484);广东省自然科学基金项目(500184)

摘　　要：随着网络技术的不断发展,入侵检测技术作为安全防护的重要手段,显得日益重要。针对现有入侵检测系统识别率低、误报率高的问题,将SOM神经网络结合Agent技术应用到入侵检测系统。结合模糊逻辑的思想对SOM网络的学习算法进行了模糊化改进,利用SOM网络的自组织特性对网络数据流量强度进行建模和聚类;使用Agent技术进行实时监控网络环境的安全状况、入侵企图的识别等。提出一个基于自组织神经网络的智能入侵检测模型,描述了模型体系结构及其工作流程。通过实验进行仿真,实验证明系统有较好的识别率和较低的误报率。With the development of Internet,as an important means,the intrusion detection technic seems necessary.In view of the problems of the existing Intrusion Detection System,namely the low recognition rate and high false rate,in this paper,the SOM neural network combined with Agent technology is introduced into Intrusion Detection System.The learning algorithm of SOM network is improved by fuzzy logical,the self-organizing characteristics of SOM network are used to model and cluster the Network data flow intensity,and using the Agent technology to supervise the safety states of the network in real time and identify the attempted invasion and so on.An Intelligent Intrusion Detection model based on self-organizing neural network is presented,and it describes the model structure and its workflow.The simulation result shows that the system is proved to have a better recognition rate and a lower false rate.

关 键 词：入侵检测系统 自组织特征映射 聚类 神经网络 移动代理 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]