Real-Time Classification Method of Network Traffic Based on Parallelized CNN

Abstract

The era of big data brings explosive network traffic. Classification of network traffic can enhance the controllability of the network, help relevant personnel to grasp the distribution of traffic on the network, help network operators optimize service quality and prevent various cybercrime. However, traditional network traffic classification methods cannot satisfy the requirements of large-scale network traffic classification, and the current machine learning-based algorithms for network traffic classification are difficult to complete the high-speed parallelized real-time traffic classification task in the actual network environment. In this paper, a method is proposed to optimize the Convolutional Neural Networks (CNN) model in parallel using the Spark platform, and the Spark Streaming framework is put forward to implement the requirements for real-time classification of network traffic. Besides, the performance of our method is evaluated. The experimental results show that the proposed method has good real-time performance while ensuring high classification accuracy, and it can implement the task of highspeed parallel network flow real-time classification.