Clustering Unknown IoT Devices in a 5G Mobile Network Security Context via Machine Learning
Tony Hämmäinen, Julen Kahles · 2021
We propose a novel machine-learning pipeline for clustering unknown IoT devices in an industrial 5G mobile-network setting. Organizing IoT devices as few homogeneous device groups improves the applicability of network-intrusion detection systems. More specifically, we develop feature engineering methods that transform IP-flows into device-level data points, define distance metrics between the data points, and apply the DBSCAN algorithm on them. Our experiments on a simulated IoT device network with varying levels of noise show that our proposed methodology outperforms alternative methods and is the only one producing a robust grouping of the IoT devices with noise present in the traffic data.