论文部分内容阅读
针对网络流量在线识别的难题,提出一种聚类算法和在线流量识别方案.以网络数据流的若干初始数据包作为子流,提取子流的统计特征,应用基于滤波器算法的属性相关性算法提取子流最佳特征子集,并提出基于密度的在线带噪声空间聚类算法对子流特征向量进行聚类,采用优势概率业务实现聚类和应用类型的映射.实验结果表明,该方案具备识别新应用类型和加密数据流的功能,且能实现在线的网络流量分类.
Aiming at the problem of on-line identification of network traffic, a clustering algorithm and online traffic identification scheme are proposed.A number of initial data packets of network data stream are used as sub-streams to extract the statistical characteristics of sub-streams, and the attribute correlation algorithm based on filter algorithm And extract the best subsets of the substreams and propose an online clustering algorithm based on the density of noise with noises to cluster the substream feature vectors and use the probabilistic probabilistic services to realize the mapping of the clustering and application types.Experimental results show that this scheme has Identify new application types and encrypt data streams, and enable online web traffic classification.