摘要
对无线网络流量异常数据信息的检测,能够有效解决计算机网络安全问题。对流量异常数据信息的检测,需要先得到网络流量数据特征间的关联性,计算各个数据对象的距离,完成异常数据信息的检测。传统方法将动态阈值作为流量行为检测点是否正常的准则,设定网络流量行为置信区间,但忽略了计算各个数据对象的距离,导致检测精度偏低。提出基于主元分析的无线网络流量异常数据信息检测方法。首先对网络流量数据输入变量进行特征提取,消除网络输入变量之间的共线性,然后对网络流量数据潜在变量进行选择,进而实现与异常数据信息检测模型输出类别最近似的异常数据特征变量的选择,得到网络流量数据特征间的关联性,计算各个数据对象的距离,利用异类挖掘理论实现无线网络流量异常行为检测。实验结果表明,所提方法能够有效提高网络流量异常检测性能,具有较好的鲁棒性。
This research focuses on the detection method for abnormal data information of wireless network traffic based on principal component analysis. The research carried out feature extraction for input variables of the network traffic data and eliminated collinearity among the input variables, then selected latent variable of the network traffic data and achieved selection of characteristic variable of the abnormal data which are the most approximate to output category of detection model of the abnormal data. The research obtained relevance between features of the traffic data and the calculated distance of each data object. Thus, using the theory of outlier mining, the detection of abnormal behavior was achieved. Experimental results show that the method can improve anomaly detection performance of net- work traffic and has better robustness.
出处
《计算机仿真》
北大核心
2017年第9期408-411,共4页
Computer Simulation
关键词
无线网络
异常数据信息
检测
Wireless network
Abnormal data information
Detection
