Deep Learning in 5G Wireless Networks - Anomaly Detections

The year of 2020 is critical for global implementation of 5G wireless networks. While enjoying a whole new level of user experience in 5G networks, such as high data rate, low latency and virtually everything to everything connections, the ever growing diversity, complexity of network and data traffics impose a set of new challenges for effectively operating and managing 5G networks. As our daily lives are more dependent on mobile devices and apps, so does the cyber security risk and venerability increase. Many of the algorithms, protocols and practices used to safeguard 4G networks fall short for 5G networks without degrading the performance expected for 5G networks. In this paper we report our early research results of using deep learning algorithms for anomaly detection in 5G network while minimizing the impacts to network latency. We developed a prototype model using U-Net and conducted a simulation experiment with a well known botnet dataset to evaluate the suitability and performance.