深度学习算法的P2P流量识别与控制  

P2P traffic identification and control based on deep learning algorithm

作　　者：黄金铃 HUANG Jin-ling(Information Technology Office,Shanghai Normal University,Shanghai 200234,China)

机构地区：[1]上海师范大学信息化办公室,上海200234

出　　处：《沈阳工业大学学报》2022年第2期209-213,共5页Journal of Shenyang University of Technology

基　　金：国家自然科学基金青年科学基金项目(61702333).

摘　　要：为了提升P2P流量的识别精度与控制效果,提出了深度学习算法的P2P流量识别与控制方法.采用P2P流量数据训练深度学习算法的BP神经网络,根据训练好的神经网络对训练样本进行预分类,得到包含各服务流量特征的预分类结果.将预分类结果作为P2P流量聚类中心值,通过聚类算法检测P2P流量样本数据,得到P2P流量识别结果.采用分形自回归综合滑动平均模型分析P2P流量控制机制.结果表明,该方法的识别性能稳定、识别结果精度较高,有效降低了流量传输的丢包率,可对P2P流量传输进行稳定控制.In order to improve the identification accuracy and control effect of P2P traffic,a P2P traffic identification and control method based on a deep learning algorithm was proposed.The BP neural network of deep learning algorithm was trained by P2P traffic data.According to the trained neural network,training samples were pre-classified to obtain the pre-classification results including the characteristics of service traffic.The pre-classification results were taken as the clustering center values of P2P traffic,the P2P traffic sample data were detected by clustering algorithm,and the P2P traffic identification results were obtained.The P2P traffic control mechanism was analyzed by a fractal autoregressive-integrated-moving-average(FARIMA)model.The results show that the identification performance of as-proposed method is stable and the accuracy of identification results is relatively higher.It can effectively reduce the packet loss rate of traffic transmission,and can stably control the P2P traffic transmission.

关 键 词：P2P流量识别 流量控制 神经网络 聚类算法 自相似模型 聚类中心 深度学习 检测样本 

分 类 号：TP393.07[自动化与计算机技术—计算机应用技术]