摘要
为了在高速网络中实时、准确地识别多种协议产生的分组流,首先分析现有的识别方法,然后从应用协议有限状态机角度出发,提出使用完美有限状态机统一描述不同识别方法的思想。在此基础上,提出了基于流序列的分组流识别方法,利用协议有限状态机中状态变化特征识别特定协议产生的分组流,并提出了一个状态特征聚合算法,以构造协议状态特征集合。通过识别基于UDP的Skype语音流,说明此方法的可行性和有效性。
It is critical to real time acquire network flow information for network management and security.In order to accurately identify various application flows in high-speed networks,a thorough analysis of the existing traffic identification methods was made firstly.And then on the perspective of protocol finite state machine,a Perfect Finite State Machine（PFSM for short） which can describe different identification methods was put forward.On this basis,a Network Flow Identification Method based on Flow Sequence（FIFS for short） was proposed,which utilizes the characteristics of PFSM states transformation to identify protocol-specific flows.One state＇s feature aggregation algorithm was proposed for constructing the set of protocol state characteristics.Finally,by identifying UDPbased Skype flow,the feasibility and availability of FIFS were verified.
出处
《解放军理工大学学报（自然科学版）》
EI
北大核心
2010年第6期628-633,共6页
Journal of PLA University of Science and Technology(Natural Science Edition)
基金
国家863计划资助项目(2007AA01Z418)
关键词
P2P
有限状态机
流识别
应用协议
机器学习
P2P
finite state machine
flow identification
application protocol
machine learning
