Abstract:Network traffic classification is a key technology for multimedia service management in smart city networks, which is of great significance to broadcast applications to smart cities. According to the characteristics of different network traffic to ensure the full utilization of broadband resources and improve the quality of network service. Due to the wide application of dynamic port number and stream encryption technology, the traditional network traffic classification method based on port number and deep packet inspection is no longer effective. In recent years, machine learning methods can be used to accurately identify categories of network traffic applications by labeled data. However, collecting enough labeled network traffic data is a time-consuming and difficult task, and the distribution of traffic data collected in various networks is different. In order to solve the problem of insufficient labeled samples and domain distribution differences, a network traffic classification method based on domain adaptive migration is proposed in this paper. This paper designs a new domain adaptive method, which reduces distribution differences between domains, makes the distribution of samples under the same category more compact and different application categories more separated. Then migrate the feature extraction module to the target domain. And the unlabeled data and few labeled samples are utilized to jointly classify network traffic applications using a siamese sparse denoising stacked autoencoder. Experimental results show that compared with other algorithms, the network traffic classification algorithm proposed in this paper achieves the best result.

Network Traffic Classification Based on Domain Adaptive Migration for Multimedia Services in Smart City Networks

Novel Dynamic Multiple Classification System for Network Traffic

Cross-domain Network Traffic Classification Using Unsupervised Domain Adaptation

A Summary of Traffic Identification Method Depended on Machine Learning

Traffic Classification - Towards Accurate Real Time Network Applications

Research on Network Traffic Classification Based on Machine Learning and Deep Learning

Composite Lightweight Traffic Classification System for Network Management

Hybrid Internet Traffic Classification Technique

Traffic Classification for Mobile Video Streaming Using Dynamic Warping Network.

Study on Internet Traffic Classification Using Machine Learning

IP Traffic Classification Based on Machine Learning

Network Traffic Classification Techniques and Comparative Analysis Using Machine Learning Algorithms

Online Internet Traffic Identification Algorithm Based on Multistage Classifier

A method of combining traffic classification and traffic prediction based on machine learning in wireless networks

Network Traffic Classification Algorithm Based on Spectral Clustering

Network traffic classification based on federated semi-supervised learning

SmoteAdaNL: a learning method for network traffic classification

FEAML: A Mobile Traffic Classification System with Feature Expansion and Autonomous Machine Learning.

A Statistical-Feature-based Approach to Internet Traffic Classification Using Machine Learning.

Network traffic grant classification based on 1DCNN-TCN-GRU hybrid model

Machine Learned Real-Time Traffic Classifiers