On the performance of a deep learning-based anomaly detection system for 5G networks

Abstract

Cyber security defense systems have evolved in the last years with innovative approaches and new Intrusion Detection Systems (IDS) capable of identifying cyber-threats that had gone unnoticed to date. However, new protection challenges are raised by the upcoming fifth generation (5G) mobile technology with new and advanced features. 5G will make that existing detection procedures become obsolete in case they are not adapted accordingly. This paper proposes a 5G-oriented architecture with which to conduct the analysis of network flows to identify cyber-threats in 5G mobile networks efficiently and quickly, by making use of deep learning techniques. Experiments on the proposed system with inspection capabilities are also included, so as to analyze and determine how many network flows, gathered from the 5G subscribers' User Equipments (UE), we can inspect in real-time. These outcomes can give us clues about when and why protection systems in 5G will stop detecting cyber-attacks for overload reasons.