Deep Learning-Based Anomaly Detection in Cyber-Physical System

Since the introduction of personal computers, software has quickly permeated every aspect of our life. When it comes to cyber-physical systems, software becomes a physical part of the system. When cyber and physical systems work together, new possibilities and difficulties arise. Cyber components offer additional levels of complexity to CPSs, which are already ubiquitous in modern civilization, with software controlling everything from automobiles and aircraft to water purification facilities, smart grids, and trains. In order to prevent costly system outages, researchers in CPS study the methods of spotting unusual patterns of behavior. In this study we present a taxonomy for anomaly detection in CPS which is an open research challenge. We discuss state of art methods used in industrial and spacecraft anomaly detection using Deep Learning models. We introduce a classification scheme for CPS anomaly detection and provide further details on the existing testbed dataset. Lastly the challenges are highlighted for further research direction.