Traffic Classification of 5G Packet Traces

Abstract

With the usage of mobile phones, privacy concerns have been a long-standing issue, and the recent advancement of 5G technology has only amplified these concerns. While encryption is a crucial method for protecting user privacy, studies indicate that machine learning techniques can identify web and mobile applications even though the traffic is encrypted.

To explore this problem, this project aims to investigate the potential for identifying mobile applications during encrypted communication on a 5G network. The project utilizes three machine learning models, namely k-Nearest Neighbors (k-NN), Random Forest, and Long Short-Term Memory (LSTM). To achieve this goal, various factors are analyzed, including the type of traffic, packet size, and timing information, to identify specific mobile applications.

This project’s results show that it is possible to identify an app over a 5G network with an accuracy of 85% approximately, raising privacy concerns on communications over a 5G network. Under this context, this job updates the current State of Art regarding the private communications over an encrypted network, showing how privacy is vulnerable in 5G networks.