Abstract

Behavioral anomalies are typically interpreted as malfunctions of an observed system. The use of machine learning offers a key advantage in anomaly detection by training on nominal system operation data only. The anomaly detection algorithms then consider deviations from this nominal behavior as faulty. In this workshop, we will primarily consider two application examples that are in the domain of anomaly detection. On the one hand, an approach based on Artificial Neural Networks will be presented that allows classification and can be used for embedded systems. On the other hand, a general data-driven approach is considered that uses various machine learning techniques to detect anomalies. In both parts, we present the levels from application to implementation.