Detection of stealthy false-data injection attacks on safety-critical cyber-physical systems

Cybersecurity solutions ensuring tight access control are essential in preventing intrusions that could inflict devastating impacts on cyber-physical systems (CPSs) operating in mission-critical areas, including electrical power grid and energy systems, traffic flow management, defence and homeland security, unmanned aerial vehicles, water treatment and distribution systems, and healthcare systems. Due to the nature of such applications, the level of reliability and security is a strong need to preserve the proper functioning of these critical systems, since their failure can cause irreparable harm to the physical system being controlled and to people who depend on it. This research report presents at first a literature review of the cyber-physical security measures designed to protect one or more facets of the CIA triad in safety-critical applications. Then, the vulnerabilities of such CPSs are investigated under a variety of cyber-attacks, ranging from disclosure attacks to data integrity and DoS attacks. Furthermore, an overview of data-driven and model-based methods for threat detection and mitigation is provided, along with the implementation of a covert attack on a DT state-space representation of a stochastic LTI system that highlights the drastic effects of cyber-physical attacks. Finally, the use of an autoencoder-based anomaly detection technique is proposed and tested on the SWaT testbed, a water treatment testbed for research and training on ICS security.