一种基于KPCA和BP神经网络的多核入侵检测分类系统的研究

Multi-kernel intrusion detection system based on KPCA and BP neural network

针对当前入侵检测系统的弱点,将KPCA技术和BP神经网络相结合,提出了一种多核入侵检测分类系统的设想.该系统针对一些复杂网络数据维数较高的特点,引入核主成分分析技术对其进行降维处理,从而简化了神经网络规模,降低了神经网络的运算量.通过对KDD99数据集进行仿真实验表明,与仅使用BP神经网络的入侵检测系统相比,该系统具有很强的泛化能力和较高的检测效率.

In view of the weakness of current intrusion detection system, a new intrusion detection system model based on the combination of KPCA technology and BP Neural Network is put forward. Against the high dimen- sions problem of complicated network data, KPCA technology as a method of characteristics extraction is used to decrease the dimensions and simplifie the size of neutral network and reduces the operations work. A large a- mount of experiments with KDD99 dataset have been conducted and the results show that the new system is with higher adaptable ability and higher speed detection rate in nowadays complicated network circumstances than the intrusion detection system only uses BP neural network.