Anomaly Detection for IoT devices using Hierarchical Self-Organizing Maps

Abstract

With the growing use of the Internet of Things (IoT), the IoT platforms, and the solutions and services related to them, cybersecurity stays of utmost importance. The cyber and interconnected nature of IoT devices makes them vulnerable to various types of cyber-attacks as well as data falsifications. In this thesis, I design and implement a data-driven unsupervised learning approach for anomaly detection for IoT devices such as smart meters, and intelligent electronic devices in power generators. To this end, I employ and evaluate Hierarchical Self-Organizing Maps on real-world smart meter and power system data collected in the USA and Europe. Results show that different types of anomalies could be detected with an F1-score between 0.857 and 0.980 based on the dataset and the type of attack observed.