摘要
针对新兴网络应用无法使用传统的基于端口与特征码进行识别的问题,对基于流量统计分析的网络协议识别方法进行了研究,提出了基于自适应BP神经网络的流量识别算法。对BP神经网络结构难以确定、易陷入局部极小值等缺陷进行了分析,使用双粒子群算法对BP神经网络进行优化以提高识别率。实验表明,该算法能根据网络流量的统计特征有效地识别网络应用,且对于采用UDP协议的应用同样有较高的识别率,同时优化后的自适应BP神经网络训练时间更短;并能自动调整其结构,具有良好的自适应特性。
Intemet traffic identification is currently an important challenge for network management. Traditional approaches focus on identifying TCP flows and cannot accurately classify emerging network applications. In this paper, a new approach based on adaptive back-propagation （BP） neural network is proposed to identify both TCP and UDP traffic flows. This approach uses the dual particle swarm optimization （PSO） algorithm to optimize the BP neural network. The experimental results show that the proposed approach can classify both TCP and UDP traffic flows at a high rate and can reduce the training time and adjust the number of hidden layer nodes of BP neural network adaptively.
出处
《电子科技大学学报》
EI
CAS
CSCD
北大核心
2012年第4期580-585,共6页
Journal of University of Electronic Science and Technology of China
基金
国家973项目(2007CB311106)
国防重点实验室基金(NEUL20090101)
关键词
自适应算法
神经网络
粒子群优化
统计特征
流量识别
adaptive algorithm
neural networks
particle swarm optimization
statistical characteristic
traffic identification
