A Real-Time Network Traffic Identifier for Open 5G/B5G Networks via Prototype Analysis

Abstract

Nowadays real-time traffic occupies lots of network resources, thus identification and analysis for this network traffic becomes urgent for operator and commercial company. With the identified traffic types, quality of experience (QoE) monitoring and optimization, user behavior analysis and network resource allocation are more beneficial for open 5G/Beyond 5G (5G/B5G) networks. Exiting studies usually adopt transport or application layer information to identify traffic, while we jointly consider them simultaneously to achieve general purpose identifier. Besides, we also analyze the flow-based features to reduce the corresponding complexity for low-complexity implementation. Based on anatomy of network traffic identification, we propose a traffic type identification framework for real-time traffic. In mainstream voice over Internet protocol (VoIP) call and video streaming services, the proposed method can achieve as much as 30% identification accuracy improvement and have more than 20% reduction in terms of the identification delay if compared with other conventional schemes.