基于递归率REC特征的网络流量相空间重构监测

发布时间：2018-04-04 19:29

本文选题：网络流量　切入点：递归率　出处：《计算机科学》2013年11期

【摘要】：传统上对网络流时间序列分析多采用线性分析方法,没有充分利用到网络流客观存在的非线性特征信息,从而使数据分析能力受限。提出了基于定量递归分析递归率REC特征的网络流量相空间重构监测模型,基于相空间重构和递归图分析,设计了网络流量的REC递归率的定量递归特征作为网络流量序列分析的数据支撑。使用平均互信息算法和虚假最近邻点算法求取流量序列的相空间重构的关键参数,利用递归图中有规律的点线检验网络总出口流量的确定性和可预测性,利用REC特征监测网络流量序列的异常流量和特性进行分析。仿真实验表明,网络流量序列的定量递归特征具有较强的稳定性和自相似性,精度较传统特征统计方法提高19%以上,采用REC递归率特征对异常流量序列的预测预报监测准确率为99.7%,比采用传统的其它非线性递归特征提高了13.2%,展示了算法在网络流量和非平稳数据序列分析中的优越性能。
[Abstract]:The traditional network flow time series analysis using linear analysis method, did not make full use of the nonlinear network flow characteristic information exists objectively, and make the data analysis ability is limited. The recurrence quantification analysis of recurrence rate characteristics of REC network traffic monitoring model based on phase space reconstruction analysis, phase space reconstruction and recursive graph based on the design. The network flow rate REC recursive recursive feature as quantitative sequence analysis of network traffic data support. The key parameters of average mutual information algorithm and false nearest neighbor algorithm for phase space reconstruction and flow sequence, a deterministic point line rule test network flow prediction and recursive graph, was analyzed by using the abnormal traffic monitoring network traffic characteristics and characteristics of REC series. Simulation results show that the quantitative recursive features of network traffic sequence has strong The stability and self similarity characteristics, accuracy than the traditional statistical method is improved by more than 19%, the accuracy rate of REC recursive feature of abnormal traffic sequence prediction was 99.7%, 13.2% higher than the other traditional nonlinear recursive feature, show superior performance of the algorithm in the analysis of network traffic and non-stationary data sequence.

【作者单位】：西南民族大学网络中心;西南民族大学计算机科学与技术学院;
【基金】：中央高校基本科研业务费专项基金项目(12NZYQN27)资助
【分类号】：TP393.06

【参考文献】