A network traffic classification framework based on deep learning

With the rapid development and widespread popularity of the Internet, the types and quantities of abnormal traffic are also increasing. Abnormal traffic detection, as an important part of the field of computer system and network security, has become a research hotspot in the information age. Traffic classification based on machine learning is one of the most widely studied abnormal traffic detection methods, but how to design and extract the relevant feature set that can achieve accurate description of traffic characteristics is the main problem faced by current research. Compared with traditional machine learning, deep learning does not require manual design and extraction of feature sets, which brings new opportunities for the development of traffic classification and anomaly detection. Therefore, this paper studies traffic classification and anomaly detection based on deep learning. The main research work and innovations are as follows: (1) Optimize and build network traffic data sets. In traffic classification and anomaly detection methods based on machine learning, data sets are the training basis of classification models, and the quality of data sets plays a key guarantee role in verifying the effectiveness of classification methods. (2) Aiming at the complexity of feature design and extraction of traditional machine learning traffic classification methods, a traffic classification method based on Net flow and DNN (Deep Neural Network) is proposed. (3) In order to solve the problem that traffic feature sets are difficult to extract in encrypted traffic classification, a Convolutional Neural Networks (CNN) and SAEN (Stacked Autoencoder Networks, Stack self-coding network), which uses a hybrid feature set construction method combining traffic statistics and spatial features. (4) In order to improve the efficiency of abnormal traffic detection, work is carried out from two aspects: first, combining abnormal traffic detection methods based on statistics; The second is to optimize the deep learning model and reduce the training and classification time of the model. A Hybrid Method of Entropy and SSAE-SVM(HESS) for abnormal flow detection is proposed.