Service identification method applicable to broadband access network

Abstract

The invention discloses a service identification method applicable to a broadband access network. Multiple data frames of different servers of the broadband access network are screened, so that data frames corresponding to TCP and UDP protocols are obtained; service flow division of the data frames is carried out according to quintuple data; service flow flow-pool groups are established; various service flow flow-pool groups each comprise an uplink service flow flow-pool and a downlink service flow flow-pool; for each service flow flow-pool, a service type label is added; then, an integrated convolutional neural network is constructed; the integrated convolutional neural network comprises an uplink one-dimensional convolutional neural network, a downlink one-dimensional convolutional neural network, an interactive one-dimensional convolutional neural network and an arbitrament module; the three one-dimensional convolutional neural networks are respectively trained by adopting the service flow flow-pool groups; when the broadband access network is practically operated, the captured data frames are standardized into service flow data packets; and the service flow data packets are input into the convolutional neural networks, so that a service identification result is obtained. According to the service identification method applicable to the broadband access network disclosed by the invention, due to introduction of a convolutional neural network algorithm in deep learning, fine-grained and high-precision service identification can be realized.

Description

technical field [0001] The invention belongs to the technical field of broadband access networks, and more specifically relates to a service identification method applied to broadband access networks. Background technique [0002] With the continuous development of broadband access network technology, broadband optical fiber access network technology has become one of the current hot technologies due to its advantages of large available bandwidth, good signal quality, and support for broadband services. Among them, the PON (Passive Optical Network, Passive Optical Network) technology has become the mainstream of today's access network technology due to its simple topology and low maintenance costs. Due to the pressure of bandwidth upgrade in PON access network technology, the processing capacity of its network equipment has developed from 1Gbps and 2.5Gbps to the current 10Gbps in just a few years, and even has a tendency to develop to 40Gbps and 100Gbps in the next few year...

AI Technical Summary

Problems solved by technology

[0010] However, the traditional DPI (Deep Packet Inspection, Deep Packet Inspection) service identification technology mainly relies on the manual extraction of service feature fields in the early stage to maintain accuracy. Although this method is effective, the preparation work in the early stage and the update and maintenance in the later stage too time consuming

The subsequent DFI (Deep Flow Inspection) business identification technology relies on a large number of extracted flow features for further screening. Using traditional machine learning classification algorithms, a large number of papers and experiments have proved the effectiveness of the method, but this reliance The empirical flow feature selection method may fail to extract or lose many key features, which greatly reduces its adaptability and portability, and the fineness of its classification is too coarse-grained to adapt to fine-grained recognition scenarios.

Images