Anomaly Detection using a Semi-Supervised Deep Learning Model on Open 5G Core Metrics during User-Equipment Registration

Examining and detecting anomalies in key metrics from core Network Function (NF) modules during User Equipment (UE) registration is vital for ensuring Quality of Service (QoS) in ever-expanding 5G networks. Existing monitoring methods often overlook these anomaly detection approaches from their data plane perspectives. In this study, we focus exclusively on the semi-supervised Long Short-Term Memory AutoEncoder (LSTM AE) model for anomaly detection. Through the analysis of CPU and memory metrics during UE registration, stress-related anomalies are pinpointed via simulations conducted on the Open 5G Core. Results show that the LSTM AE model achieves a remarkable accuracy of over 97%, offering superior anomaly detection performance with fewer false positives. This work significantly enhances anomaly detection within core 5G network modules, ultimately contributing to improved QoS.