Spam SMS Detection from 5G Mobile Core Network Using Machine Learning-based Federated Learning

Abstract

The widespread use of SMS as a communication method emphasizes its crucial role in contemporary society. SMS offers a convenient and reliable means of transmitting information across various demographics and contexts for both personal communication and business transactions. Despite its numerous benefits, the increase in spam SMS coverage presents a significant challenge, affecting the user experience and network reliability. With the growth of IoT and the widespread adoption of 5G technology, the volume of SMS traffic is expected to rise further, exacerbating this issue. To address this problem and enhance user satisfaction, innovative solutions are necessary. This paper proposes a novel approach aimed at reducing spam SMS messages within a 5G core network, thereby minimizing unwanted messages for end users. The proposed solution combines Machine Learning (ML) and Federated Learning (FL) techniques for spam SMS detection, prioritizing user privacy and encouraging user participation for advanced anti-spam systems building. Extensive simulations using real-world examples of spam and ham have shown promising results with an accuracy of 99.12% and a detection time of 0.009ms.