Classifying 5G Encrypted Packet Traces

Paper i proceeding, 2024

In today's digital era, privacy concerns have raised and the further deployment of 5G technology in recent years has amplified some of these concerns. While encryption is undoubtedly the cornerstone for protecting user privacy, packet trace analysis has been shown to be capable of identifying web and mobile applications even when the network traffic is encrypted. We further extend previous studies by presenting an accurate traffic classifier for 5G packet traces that does not need any access to the network. Our novel classification method is based exclusively on traffic patterns and does not require any information from the network layer to function. We perform a preliminary experimental comparison study for the classification task with three machine learning models and eight popular mobile applications. Our best model reaches an accuracy of up to 87% in classifying one minute of captured 5G encrypted traffic. This prompts us to question if necessary measures need to be considered at the link layer so to alleviate to leak such information.