Artificial intelligence-based anomaly detection for the efficient management and security of the future cellular networks

Abstract

Existing communication systems are getting more intricate and a morass to manage which is galvanized by ever-increasing network capacity demand, device density, and the data traffic. Network operators in the United States alone deplete over USD 15 billion yearly to handle cellular outages, incurring escalated operational expenditure (OPEX) as a result. Furthermore, congestion in a cell degrades subscriber quality-of-experience and quality-of-service which results in an increased churn rate and subsequently reduced operator's revenue. Besides the management aspect, the security of the cellular network is paramount to prevent cyber-attacks against its infrastructure not just for its primary subscribers but also for the fact that the networks can be exploited as a proxy to attack the connected cyber-physical systems (CPSs). To unlock the full extent of 6G networks, artificial intelligence (AI)-empowerment has paramount potential for efficient network management and preventing cyber-attacks against their infrastructure. Based on this backdrop, this thesis applies AI techniques to achieve a primary objective of an efficient, scalable, and timely detection of outages and the situation leading towards congestion in a cell under the context of cellular network management. It also aims to achieve a secondary objective of detecting various cyber-attacks towards the availability of cellular network services in the context of cyber-security of cellular infrastructure and CPSs. Cell outages, situation leading towards congestion in a cell, and cyber-attacks are treated as anomalies in this work and various machine learning (subset of AI) models and data analytic tools are utilized to detect them by leveraging real subscriber data based on call detail records (CDRs) extracted from a 4G LTE-A network.