On Real-Time Traffic Classification in 5G Networks

Real-time traffic classification is crucial for enhanced performance of emerging 5G services. This motivates us to classify traffic to different services through statistical based features. We propose to use unencrypted statistical information in Downlink Control Information (DCI) messages and apply an unsupervised Gaussian Mixture Models (GMMs) for modeling individual mobile services. The proposed method is utilized for two main purposes: firstly, to classify unlabeled mobile traffic into specific services and secondly, for traffic verification to confirm if the claimed source for the traffic matches the expected service.