Detecting Intrusive Behaviour of People in a Building through Data Analysis and Anomaly Detection in Home Automation Systems
Thomas Mundt, Simeon Wiedenmann, Johannes Goltz, Johann Bauer, Maximilian Jung · 2019
Building automation systems provide a lot of data about the current condition of houses. For this purpose, a large number of sensors are usually installed in the building. Such sensors are for example motion detectors and light switches. The sensors detect human actions. They are connected to a network, in our case via a fieldbus network, namely KNX. This makes it possible to evaluate the events at a central location. In this paper we show how data from building automation can be used to detect physical intruders and other "anomalies" such as unusual human movements. We present how we collected data about the physical location of the sensors and data representing events. We also show what procedures and algorithms we have used to detect anomalies.