A method of traffic identification based on bp neural network

Abstract

The invention discloses a BP-neural-network-based method for performing traffic identification. The method comprises the steps that characteristic values are generated by obtaining simple packet header information, and the suitable characteristic values are selected from the generated characteristic values as input of a BP neural network to obtain a sample training set; afterwards, network data streams are captured in real time, and characteristics are extracted to serve as the input of the BP neural network for real-time identification. The extracted characteristic values based on a time window method have extendibility and are easy to obtain. The three-layer neural network is selected as the implementation scheme, and initialized weight values having global optimal characteristics are sought based on particle swarm optimization (PSO); a distributed and parallel netty structure is adopted for the BP neural network for training and learning, so that the fault tolerance of the BP neural network is higher, the processing speed is higher, and the real-time identification can be better completed. Meanwhile, as the BP neural network is trained in a global optimization mode, the BP neural network has extremely high generalization ability and can deal with more complex and diversified network environments in the future.

Description

technical field [0001] The invention relates to the field of network traffic identification, in particular to a method for traffic identification based on a neural network. Background technique [0002] With the advent of the information network era, network data has shown explosive growth, and there are more and more network applications, followed by users' higher and higher requirements for network bandwidth. This results in insufficient network bandwidth and increased network congestion. Network traffic identification can play an important role in improving safe, reliable and high-quality services for users. Therefore, network traffic identification and QOS control management have received more and more attention. [0003] Existing network traffic identification methods are mainly divided into four aspects: traffic identification methods based on port numbers, traffic identification methods based on payload characteristics, traffic identification methods based on network...

AI Technical Summary

Problems solved by technology

However, the current identification method of machine learning is not real-time, and cannot identify the current user's network usage in real time.

Analysis by capturing data packets for a long time is not only complex but also difficult to implement

With the rapid development of the network, the data situation of the network will become more complex and diverse. The current machine learning is not conducive to expansion and does not have good generalization ability

Images