Traffic Classification Based on CNN-LSTM Hybrid Network

Abstract

AbstractNetwork traffic classification is a significant method of network anomaly detection and plays a critical role in cyberspace security. Accompanied by the rapid development of network traffic diversity, traditional detection methods are not any longer adapt to the complex network environments. In the article, we use the method of deep learning and present a traffic classification method, which directly operates on raw traffic data. A hybrid neural network combining 1D CNN and LSTM network is used for learning the spatial and temporal characteristics of the stream in the meantime, which is verified on the CICIDS2017 dataset. Finally, we applied the model verified on the public data set to real network traffic, and achieved good results. It can be concluded from the experimental results that our arithmetic has higher accuracy on public dataset as well as can be applied to real network environments to solve real-world problems.KeywordsTraffic classificationDeep hybrid networkRaw featureReal network environment