大数据环境下光纤通信网络异常流量动态检测方法

Dynamic detection method of abnormal traffic in optical fiber communication network under big data environment

大数据环境下光纤通信网络具有多样性特征,增加了网络异常流量检测的难度。为提高异常流量检测的性能,提出一种基于指数平滑和置信区间的异常流量动态检测方法。随机划分连续时间光纤网络数据,利用二叉分割组建决策树模型,计算数据差异分数后保留完整网络流量数据。推算各流量特征密度指标,使用高阶统计量推导可能产生异常流量的时间范围。通过滑动窗口模型获得近期相同时段流量值,融合中心极限定理分析流量是否位于临界值上下置信区间,输出异常流量检测结果。实验结果表明,所提方法检测正确率高、计算速率快、稳定性强。

Optical fiber communication networks in the big data environment have diverse characteristics,which increase the difficulty of detecting abnormal network traffic.In order to improve the performance of anomaly flow detection,a dynamic anomaly flow detection method based on exponential smoothing and confidence interval is proposed.The data of continuous-time fiber optic networks are randomly partitioned,and the decision tree model is established by binary segmentation.The characteristic density index of each flow is calculated,and the time range of abnormal flow is deduced by using high-order statistics.Using the sliding window model to get the same traffic value in recent period,the fusion center limit theorem analyzes whether the traffic is in the confidence interval above and below the critical value.Experimental results show that the proposed method has high detection accuracy,fast calculation rate and strong stability.