一种基于Chebyshev的网络流量异常检测方法  

A NETWORK TRAFFIC ANOMALY DETECTION METHOD BASED ON CHEBYSHEV

作　　者：胡平[1] 叶坤[1] 刘瑞琴[1] 

机构地区：[1]南京工业大学电子与信息工程学院,江苏南京211816

出　　处：《计算机应用与软件》2016年第5期127-131,共5页Computer Applications and Software

基　　金：连云港市科技项目(SH1110)

摘　　要：网络安全问题已日趋频繁出现在人们的生产与生活当中,并越来越受到重视。为了能及时有效地预警网络异常,提出一种基于网络流量的异常检测方法,并针对流量数据的特性而采用时间序列的检测方式。在滑动窗口内先对序列进行格拉布斯坏点数据预处理,再利用欧式距离提前判定和切比雪夫多项式系数判定相结合方法,对其进行快速异常检测。仿真实验分析表明:在一定条件下,该算法在保证较好的检测精度的前提下,仍具有较快的检测速度,可以满足实时检测的需要。Network security problem has been occurred in people＇s production and life more frequently,and has attracted growing attention. For the purpose of effective forewarning the anomaly in networks,this paper proposes a network traffic-based anomaly detection algorithm. The algorithm adopts detection mode in time series for the characteristic of traffic data,and uses Grubbs criterion to preprocess the outlier data of the series within each sliding window first,then makes fast anomaly detection on it using the method of combining the earlyjudging with Euclidean distance and the judge with Chebyshev polynomials coefficient. The analysis of simulative experiment demonstrates that under certain conditions,the proposed algorithm has a faster detection speed while keeping preferable detection precision. It can meet the demands of real-time detection.

关 键 词：网络流量数据 异常数据 切比雪夫多项式 时间序列 格拉布斯准则 

分 类 号：TP301[自动化与计算机技术—计算机系统结构]