基于并行特征选择和分类的网络入侵检测方法

针对存在大量访问时的网络入侵检测问题,提出一种在MapReduce框架下实现的并行网络入侵检测方法。构建一种并行化的量子粒子群优化(QPSO)算法,对原始数据集中的大量特征进行选择,降低特征维度;实现一种并行化的朴素贝叶斯(NB)分类器,以网络访问特征作为输入来检测入侵。在KDDCup99数据集上的实验结果表明,该特征选择方法能够选择出最优特征子集,有效提高了入侵检测的准确性,特征选择和分类器的并行化缩短了检测时间。

Network intrusion detection based on parallel feature selection and classification

To solve the problem of network intrusion detection when there is a large number of accesses, a parallel network intrusion detection method based on MapReduce was proposed. A parallel quantum particle swarm optimization (QPSO) algorithm was constructed to reduce the feature dimension by selecting feature subset from the original data set. A parallel naive Bayesian (NB) classifier was implemented in which network access features were taken as input to detect intrusions. Experimental results on KDDCup99 dataset show that the proposed feature selection method can select the optimal feature subset and improve the accuracy of intrusion detection effectively. The parallelization of feature selection and classifier greatly reduces the detection time.