基于机器学习与DBN网络的网络入侵检测方法研究

Research on Network Intrusion Detection Method Based on Machine Learning and DBN Network

随着计算机网络的发展,网络入侵的情况也越来越严重。传统网络入侵检测方法存在检测效率低、误判率高的情况,为了解决这些问题,提出了一种基于支持向量机的深度置信网络(SVM-DBN)的入侵检测方法。通过对支持向量机(SVM)进行优化,将支持向量机与深度信念网络(DBN)融合,利用SVM、DBN与SVM-DBN在网络入侵数据集中进行对比。结果表明,SVM-DBN算法的误差率最低,比DBN和SVM的误差率平均值分别低了8.95%,12.70%,且SVM-DBN算法在训练次数为140次时最大绝对百分比误差为4.8%,均优于对比方法。这说明SVM-DBN网络能够有效地提高网络入侵检测的精度和效率。

With the rapid development of computer networks,network intrusion has become more and more serious,and the traditional network intrusion detection methods have low detection efficiency and high false positive rate.In order to solve these problems,the research proposes an intrusion detection method based on support vector machine-deep belief network(SVM-DBN).By optimizing support vector machine(SVM),SVM-deep belief network(DBN)is fused.SVM,DBN and SVM-DBN are used to compare in network intrusion dataset.The results show that the SVM-DBN algorithm has the lowest error rate,which is 8.95%and 12.70%lower than the average of the error rates of DBN and SVM,respectively,and the SVM-DBN algorithm has a maximum absolute percentage error of 4.8%when the number of training times is 140,which is better than the comparison methods.This indicates that SVM-DBN network can effectively improve the accuracy and efficiency of network intrusion detection.