基于高阶统计特征的大数据异常负载检测仿真  

Simulation of Big Data Abnormal Load Detection Based on High Order Statistical Characteristics

作　　者：兰瑞乐 唐忠 刘晓红 LAN Rui-le;ANG Zhong;LIU Xiao-hong(Network Information Center of Nanning Normal University,Nanning Guangxi 530001,China;College of Humanities and Social Sciences,Guangxi Medical University,Nanning Guangxi 530021,China;Information center of Guangxi Medical University,Nanning Guangxi 530021,China)

机构地区：[1]南宁师范大学网络信息中心,广西南宁530001 [2]广西医科大学人文社会科学学院,广西南宁530021 [3]广西医科大学信息中心,广西南宁530021

出　　处：《计算机仿真》2021年第9期329-333,共5页Computer Simulation

基　　金：广西重点研发计划(桂科AB18126068);广西教育科学“十二五”教育规划课题(2015ZKY12);广西教育研究生课程建设项目(YJSA2019005)。

摘　　要：针对大数据负载分析异常负载检测精度较低、流量耗费较高的问题,研究基于高阶统计特征的大数据异常检测方法。利用双谱值检测方法提取大数据负载中的高阶统计特征,构建高阶统计特征集合;利用人工免疫理论构建大数据异常负载检测器,以高阶统计特征集合对正常大数据负载样本编码生成集合,结合随机生成过程和高亲和力检测器克隆突变后代形成一个不成熟检测器。阴性选择算法用于将集合中的元素与未成熟检测器一一匹配,并通过低于匹配阈值的免疫耐受将它们转换为成熟的检测器,检测大数据的异常负载。仿真结果显示该方法检测结果准确率达到99.9%,耗费流量显著降低。Aiming at the problems of low accuracy and high traffic consumption of abnormal load detection in big data load analysis,a big data anomaly detection method based on high-order statistical characteristics was studied.Based on bi-spectral value detection method,high-order statistical features in big data load were extracted to build a feature set of high-order statistics.Artificial immune theory was applied to build load detector of big data anomaly.The load samples of normal big data were encoded to generate a set via the high-order statistical feature set.According to the random generation process and high affinity detector,the mutant offspring were cloned to form an immature detector.Negative selection algorithm was used to match the elements in the set with immature detectors one by one,and transformed them into mature detectors by immune tolerance below the matching threshold,detecting the abnormal load of big data.The simulation results show that this method has high detection accuracy(99.9%)and low traffic consumption.

关 键 词：高阶统计特征 大数据 异常负载 实值编码 克隆变异 

分 类 号：TP309.2[自动化与计算机技术—计算机系统结构]