网络入侵特征优化检测方法仿真

发布时间：2018-07-11 11:03

本文选题：网络入侵 + 特征检测　；参考：《计算机仿真》2013年09期

【摘要】：网络入侵特征的伪装程度越来越高,使得入侵特征与正常数据特征在分类效果上的倾斜度越来越接近。传统的采用特征分类的入侵检测方法对训练入侵特征数据集的最佳类分布未知,都是假定误分类代价均等,只注重分类的精度敏感,忽视了类型间的区别,造成入侵检测不准。提出了一种敏感性数据挖掘的网络入侵特征检测算法。利用主成分分析方法,提取网络操作数据中的主成分,去除冗余数据,将网络入侵特征的敏感性引入到检测过程中,利用敏感性数据挖掘方法,获取网络操作数据中的恶意入侵操作行为的特征,从而完成网络入侵特征检测。实验结果表明,利用改进算法进行网络入侵特征优化检测,能够准确获取网络操作行为中的异常特征。
[Abstract]:The camouflage degree of network intrusion features is higher and higher, which makes intrusion features and normal data features more and more close to the classification effect. The traditional intrusion detection method based on feature classification is unknown to the optimal class distribution of the training intrusion feature data set. It is assumed that the cost of misclassification is equal, only the accuracy of classification is sensitive, and the difference between types is ignored, which results in the uncertainty of intrusion detection. A network intrusion detection algorithm based on sensitive data mining is proposed. Principal component analysis (PCA) is used to extract the principal components from the network operation data, remove redundant data, introduce the sensitivity of network intrusion features into the detection process, and use the sensitivity data mining method. The characteristics of malicious intrusion behavior in network operation data are obtained, and the detection of network intrusion characteristics is completed. The experimental results show that using the improved algorithm to optimize the detection of network intrusion features can accurately obtain the abnormal features in the network operation behavior.
【作者单位】：东北师范大学人文学院;
【分类号】：TP393.08

【参考文献】