Implementation of Programmable CPS Testbed for Anomaly Detection

A large number of studies have provided datasets in the field of CPS security research, but the rate of actually using these datasets is low. It is difficult to objectively compare and analyze research results based on different testbeds or datasets. Our goal is to create public datasets for CPS security researchers working for anomaly detection. It is challenging for individuals to collect long-term datasets repeatedly for a large number of scenarios. This can lead to mistakes and inaccurate information. The process of collection must be comfortable and automated. For this purpose, we constructed a testbed in which three physical control systems (GE turbine, Emerson boiler, and FESTO water treatment system) can be combined with each other through the dSPACE Hardware-in-the-loop(HIL) simulator. We have built an environment that can automatically control each sensor and control point remotely. Using this environment, it is possible to collect datasets while repeatedly driving a large number of benign/malicious scenarios for a long period while minimizing human effort. We will develop and release CPS datasets using the testbed in the future.