DEEP LEARNING FOR REMOTE STATE ESTIMATION UNDER PHYSICAL LAYER ATTACKS

Abstract

With the rapid developments in sensing and communication technologies, cyber-physical systems (CPSs) have been widely applied in various engineering fields, playing a crucial role in the era of Industry 4.0. CPSs typically integrate spatially distributed plants, sensors, machines, and controllers to achieve the desired performance of physical processes. However, the integration of communication networks in CPSs makes them susceptible to malicious attacks, such as denial of service (DoS) attacks. These attackers not only disrupt communication networks but also pose significant threats to physical systems. Therefore, it is essential to design efficient control approaches to ensure the security of the CPSs in the presence of such attacks. Conventional security systems often lack effectiveness against these sophisticated attackers, whereas machine learning (ML) techniques have shown great promise in various cyber-security applications. This thesis focuses on addressing remote state estimation problems under physical layer DoS attacks using deep learning algorithms.