A Deep Learning-based Malware Traffic Classifier for 5G Networks Employing Protocol-Agnostic and PCAP-to-Embeddings Techniques

Abstract

As 5G networks become more complex, cyber attacks targeting IoT devices are deemed a serious concern. This work proposes a novel approach to detect 5G malware traffic using a network packet preprocess toolkit and machine learning models. The system can transform packets into images or embeddings, which allows for more accurate representations that can be applied in a commercial Intrusion Detection System application in a protocol agnostic manner. The paper introduces Long Short-Term Memory Autoencoders as the preprocessing method for embeddings generation followed by a Fully-Connected network for classification purposes of a 5G-dedicated dataset. The proposed approach is efficient and adaptable to evolving threats and protocols, achieving enhanced accuracy rates in detecting 5G malware traffic. This new method can facilitate defending against 5G malware attacks and paves the way for future developments in 6G networks.