摘要
网络入侵数据常常体现为高维、线性不可分性。RBF神经网络没有降维处理的功能,所以直接对原始数据进行检测速度相当慢,影响网络入侵检测的实时性。如果采用传统的选择性删除法进行降维处理,会造成信息的丢失,影响网络入侵的检测精度。为了提高网络入侵检测率和检测速度,提出一种主成分分析(PCA)和RBF神经网络相结合的网络入侵检测方法(PCA-RBF)。PCA-RBF在通过PCA对网络入侵原始数据进行维数和消除冗余信息处理的基础上,构建RBF神经网络入侵检测模型。仿真结果表明,相对于传统的RBF方法,PCA-RBF降低了漏检率、误检率、检测时间,提高了检测正确率,具有良好的检测性能。
Network intrusion data are often indicated by high-dimension and the inseparability of linear. RBF neural network does not drop the dimension of the function, which directly detects the original data and moves very slowly, and leads to the inefficiency of the real--time network intrusion detection. To reduce the dimension by traditional selective deletion will result in the loss of information and degrade the detection accuracy of network intrusion. In order to improve the network intrusion detection rate and detection speed, we are putting forward the network intrusion detection method（PCA-RBF） which combines a principal component analysis （PCA） and the RBF neural network. PCA-RBF builds the RBF neural network intrusion detection model on the base of PCA network intrusion＇ s eliminating the redundant information and processing original data dimension. The simulation results show that compared to traditional RBF method, the PCA-RBF reduces the undetected rate, the false detection rate, shortens detection time and improves the detection accuracy rate, which demonstrates its good detection performance.
出处
《计算机安全》
2013年第2期27-30,共4页
Network & Computer Security
基金
广西教育厅科研项目(201204LX313)
关键词
主成分分析
径向基函数神经网络
网络入侵
Principal component analysis（PCA）
Kadial basis function（RBF）
Intrusion detection
作者简介
张旭（1975-），男（汉族），广西桂平市人工程师，硕士，主要研究方向：数据挖掘，信息安全。
