Unsupervised Deep Learning for Behavioral and Quality Anomaly Detection in 5G Networks: Early Methods and Challenges

Next-generation wireless technology like 5G networks, Internet of Things (IoT), and cloud computing are evolving continuously and becoming a part of our daily lives. These raised security concerns, which led to a problem of managing and protecting large amounts of valuable data. Clustering, anomaly detection algorithms, and autoencoders are good for dealing with high-dimensional, unlabeled, and real-world datasets. In this study, we analyze the models that have unsupervised deep learning techniques for detecting anomalies in networking datasets. Then, we performed a pilot experiment on a public dataset named “Smartphone Usage and Behavioral Dataset” to detect behavioral anomalies among mobile users. We used unsupervised deep learning techniques, including clustering and autoencoders, to detect this behavioral anomaly and obtained the initial results discussed in the Results section. Another experiment is performed on “5G Resource Allocation Dataset: Optimizing Band” to detect resource allocation anomalies where users with inferior signal strength and high latency still receive vast amounts of bandwidth, which is unusual and inefficient.