链路失衡干扰下网络流量异常点挖掘仿真
Indoor Positioning Method of Modular Educational Robot Based on Wireless Control Technology
岑俊杰 1李永波2
作者信息
- 1. 河南工学院计算机科学与技术学院,河南 新乡 453002
- 2. 河南师范大学计算机与信息工程学院,河南 新乡 453002
- 折叠
摘要
网络数据中心流量的异常会占用大量的带宽资源,且数据流量具有多样化特征,尤其当网络链路流量失衡时,通过部署监测软件获取的节点信息已经无法实时监控流量状态.为此提出链路失衡干扰下网络流量异常点挖掘方法.利用卷积神经网络自编码器对网络流量去噪,有效控制链路失衡对流量数据挖掘的影响.通过对比正常流量点与异常流量点提取网络流量特征,结合马氏距离到改进的自编码神经网络系统中挖掘网络流量异常点.实验结果表明,研究方法的网络流量异常点挖掘准确率可稳定在 90%以上,F1 值始终高于0.8,误报率不高于 0.5%.
Abstract
Abnormal traffic in network data centers may occupy a lot of bandwidth resources.Due to the diversi-fied characteristics of data,it is unable to monitor the traffic status in real-time only according to the node information obtained by monitoring software when the network link traffic is unbalanced.For this reason,this article puts forward a method of mining network traffic outliers under link imbalance interference.Firstly,a self-encoder of a convolutional neural network was used to remove the noise in network traffic and effectively control the impact of link imbalance on traffic data mining.After the normal traffic points were compared with abnormal traffic points,the network traffic char-acteristics were extracted,and the abnormal traffic points were mined in the improved selfcoding neural network system combined with Markov distance.Experimental results prove that after the proposed method is applied,the ac-curacy of network traffic outliers mining can be more than 90%,and the F1 value is always higher than 0.8.Mean-while,the false positive rate is less than 0.5%.
关键词
链路干扰/去噪处理/网络流量特征提取/异常点检测/马氏距离Key words
Link interference/Denoising/Network traffic feature extraction/Outlier detection/Mahalanobis dis-tance引用本文复制引用
基金项目
河南省教育厅科学技术研究项目(18B880003)
河南工学院教学改革项目(2017-YB008)
出版年
2024
NETL
NSTL
万方数据