Support vector machine based dedicated network flow classification method

Abstract

The invention relates to a support vector machine based dedicated network flow classification method. The support vector machine based dedicated network flow classification method comprises the steps of grabbing independent network packets from a network and distinguishing the network packets into independent streams according to features; performing feature extraction on each stream, and describing each stream by adopting a vector form; after classifying each type of the stream, generating a sample set, then performing a support vector machine algorithm based on the sample set, classifying the samples, mapping vectors to a high dimension by adopting a radial basis function so as to realize binary classification of the samples by the support vector machine, and extending from the dichotomy to multi-class classification by using a 1-to-N extension method. The support vector machine based dedicated network flow classification method solves the problems that a dedicated network is connectionless-protocol-oriented, the current flow classification algorithm is low in adaptability and the multi-class extension is big in calculation quantity and low in accuracy, and has a very good popularization and application foreground.

Description

technical field [0001] The invention belongs to the field of network measurement and analysis, and in particular relates to a special network flow classification method based on a support vector machine. Background technique [0002] With the continuous development of network technology, the rapid increase of network bandwidth and the rapid increase of various applications in the network, new application categories are added to the network every day, and the demand for network management and network analysis is getting higher and higher. As the basis of network behavior analysis, network traffic classification and identification technology has been widely used not only in network management and analysis, but also in network security and network service quality evaluation. [0003] Network traffic identification and classification involves all aspects of business behavior. Each type of network traffic has its own behavioral characteristics. With the continuous emergence of ne...

AI Technical Summary

Problems solved by technology

However, for some private networks, such as sensor networks, military networks, emergency relief networks, etc., due to reasons such as communication channel bandwidth, reliability, and business characteristics, most of the traffic in the network is based on the UDP protocol, and the applications carried by the upper layer Layer protocols are also mostly dedicated protocols, most of which have no public standards. At the same time, because users in the network are independent of each other, there is a lack of unified coordination and planning between them, resulting in the lack of exclusivity of the protocols, that is, different protocols may appear in the application layer identification field, port number, etc. In the case of reuse, at the same time, the specificity of these networks leads to certain confidentiality requirements for the transmission content, and some encryption algorithms are usually used to protect the transmission content

All of the above properties lead to the inability of the first three classification methods to be used in these networks

Images