基于改进灰色聚类算法的云架构数据中心网络异常流量过滤算法

Cloud architecture data center network abnormal traffic filtering algorithm based on improved grey clustering algorithm

为避免异常流量影响云架构数据中心网络安全运行,需要对云架构数据中心网络异常流量进行过滤。异常流量在不同信噪比和信道条件下过滤难度不同,为了在不同过滤条件下保障异常流量过滤效果,提出了基于改进灰色聚类算法的云架构数据中心网络异常流量过滤算法。通过时间-频率分析构建了云架构数据中心网络流量传输模型,采集网络流量序列;引入加权广义距离改进灰色聚类算法,利用改进的灰色聚类算法计算网络流量序列特征最佳聚类结果,实现流量序列特征提取;通过主成分分析法获取流量序列特征的主分量特征值,构建两个子空间,将流量特征以矩阵方式映射到两个子空间中;根据映射周期向量的平方预测误差与阈值计算结果,过滤异常流量。实验结果表明,该算法可通过聚类实现数据中心网络流量序列特征提取,在不同信噪比和信道条件下有效过滤异常流量;当网络信噪比为25dB且流量在高斯信道中传输时,异常流量过滤效果更突出。

To avoid abnormal traffic affecting the safe operation of the cloud architecture data center network,it was necessary to filter the abnormal traffic of the cloud architecture data center network.The difficulty of filtering ab-normal traffic varies under different signal-to-noise ratios and channel conditions.In order to ensure the filtering effect of abnormal traffic under different filtering conditions,a cloud architecture data center network abnormal traf-fic filtering algorithm based on improved grey clustering algorithm was proposed.A network traffic transmission model was built for cloud architecture data centers through time-frequency analysis,and network traffic sequences were collected.Weighted generalized distance was introduced to improve the grey clustering algorithm,and the im-proved grey clustering algorithm was used to calculate the optimal clustering results of network traffic sequence features,achieving traffic sequence feature extraction.The principal component eigenvalues of traffic sequence features were ob-tained through principal component analysis,two subspaces were constructed,and traffic features were mapped in a ma-trix manner to the two subspaces.Abnormal traffic was filtered based on the square prediction error of the mapping pe-riod vector and the threshold calculation results.The experimental results show that this algorithm can achieve feature extraction of data center network traffic sequences through clustering,effectively filtering abnormal traffic under differ-ent signal-to-noise ratios and channel conditions.When the signal-to-noise ratio of the network was 25 dB and the traffic was transmitted in a Gaussian channel,the filtering effect of abnormal traffic was more prominent.