基于SPCSE与WKELM的网络入侵检测方法研究  

Research on Network Intrusion Detection Method Based on SPCSE and WKELM

作　　者：肖耿毅[1] XIAO Geng-yi(Dept.of Mathematics and Computer Technology,Guilin Normal College,Guilin Guangxi 541199,China)

机构地区：[1]桂林师范高等专科学校数学与计算机技术系,广西桂林541199

出　　处：《计算机仿真》2022年第6期425-429,共5页Computer Simulation

基　　金：广西教育厅自然科学基金项目(2013YB286)。

摘　　要：网络入侵检测系统是用于保护系统免受非法攻击的重要网络防御工具,网络入侵行为复杂的、冗长的特征严重影响网络入侵的检测效果。现提出一种基于稀疏主成分空间嵌入与加权核极限学习机的网络入侵检测方法。为了有效约简网络入侵数据的特征,提出一种基于稀疏主成分的特征约简的降维算法,即稀疏主成分空间嵌入算法(SPCSE)。同时,为了改进极限学习机的检测效果,提出一种加权核极限学习机算法(WKELM),它采用核函数代替包含激活函数的加权极限学习机隐层随机特征映射,有利于提高算法的非线性处理能力和鲁棒性。仿真结果显示加权核极限学习机对于网络入侵的检测精度95%,极限学习机对于网络入侵的检测精度92%,而提出的网络入侵检测方法对于网络入侵的检测精度达到98.5%,表明提出的网络入侵检测方法对于网络入侵的检测结果优于传统极限学习机以及加权核极限学习机。Network intrusion detection system is an important network defense tool to protect the system from illegal attacks, and the complex and lengthy characteristics of network intrusion bring severe challenges to the construction of effective detection system. Therefore, this paper proposes a network intrusion detection method based on sparse principal component space embedding and weighted kernel extreme learning machine. In order to effectively reduce the features of network intrusion data, a dimension reduction algorithm based on sparse principal component, sparse principal component space embedding algorithm(SPCSE), is proposed. At the same time, in order to improve the detection effect of extreme learning machine, a weighted kernel extreme learning machine algorithm(WKELM)is proposed, which uses kernel function to replace the hidden layer random feature mapping of weighted extreme learning machine including activation function, which is conducive to improve the nonlinear processing ability and robustness of the algorithm. The experimental simulation results show that the detection accuracy of weighted kernel limit learning machine for network intrusion is 95%, the detection accuracy of limit learning machine for network intrusion is 92%, and the detection accuracy of the network intrusion detection method proposed in this paper is 98.5%. The results show that the network intrusion detection method proposed in this paper is better than the traditional limit learning machine and weighted kernel limit learning machine.

关 键 词：稀疏主成分空间嵌入 加权核极限学习机 网络入侵检测 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]