Research And Implementation Of Network Application Classification And Recognition Method For Application Driven Networks

With the rapid development of network technology,the Internet is speeding up its integration with various fields of production and life.This trend has also led to the influx of a large number of new applications in the network.Traditional networks need to cope with the more complex and changeable network resource requirements of network applications and provide differentiated communication services for different applications.However,the closed and rigid architecture of traditional networks limits their ability to provide differentiated services.SDN technology can achieve flexible control of network resources and solve the problem of traditional network closure and rigidity.Under this technical background,in order to provide differentiated network services for applications,the application-driven network of "on demand" has attracted the attention of researchers in the industry.The premise of realizing "on demand" is to identify the different requirements of different applications for network resources.However,most of the existing network application classification and recognition methods are aimed at identifying protocol types and security attributes,which can not reflect the network resource requirements of applications.Therefore,from the perspective of application demand for network resources,this paper studies the classification and identification methods of application-driven network applications.The main work includes the following aspects:(1)Aiming at the problem that the current application classification methods can not describe the network resource requirements of various applications,this paper proposes an application classification method based on network resource requirements,which can cluster according to the distribution characteristics of the application network resource requirements samples,and classify the network applications with similar network resource demand into one group,and quantify the demand for network resources for different types of network applications.(2)Aiming at the problem that existing application identification methods can not identify the network resource requirements of applications,this paper combines deep learning with semi-supervised learning,and proposes a network application resource requirements identification method based on traffic characteristics.In order to speed up the construction of weak classifiers,a multi-level residual fully connected neural network model is proposed.Based on the network traffic characteristic samples marked by the first part of the research results,a weak classifier cluster is constructed and trained.On this basis,an incremental learning method based on type recognition accuracy is proposed to update the weak classifier and improve the recognition accuracy of network application resource requirements.(3)In order to meet the requirement of fast data packet capture and processing in thecurrent high-speed network environment,this paper designs and implements a network application resource requirement identification system based on DPDK technology.DPDK technology is used to extract network traffic characteristics around the Linux kernel to realize the rapid identification of network application resource requirements.In summary,in order to realize the application-driven network of "on demand",this paper studies the classification and identification method of network application based on network resource demand,proposes the application classification method based on network resource demand and the identification method of network application resource demand based on traffic characteristics,and designs and implements a set of network application resource demand identification based on DPDK technology.The feasibility and validity of the research results are verified by a series of simulation experiments and system experiments.