Perception vs. reality

Abstract

Modern cyber systems are breaking the barriers around them and is rushing to the domain of physical world. This novel type of cyber systems across the two world, or cyber-physical systems inevitably comprises interfaces between the cyber and physical world. Sensors, one of two core components of the cyber-physical interfaces, are essential for cyber-physical systems to comprehend physical world. Although sensors can offer massive benefits, the adverse effect of blindly trusting them has long been underestimated

attackers can compromise the whole system by exposing vulnerable sensors to falsified physical signals. These kinds of threats are called sensor attacks, and have recently been studied by multiple studies. This thesis comprehensively explores the field of sensor attacks and defenses. It begins with presenting multiple examples of sensor attacks against real systems or sensor modules. These examples experimentally demonstrate how sensors and the systems behind can be compromised by external attackers, which suggests sensor attack is not just a theoretical threat but rather an urgent problem to be resolved. It then discusses the difficulties of devising robust and generalizable defenses against the threat of sensor attacks by introducing research on bypassing formerly proposed sensor defense scheme. Lastly, despite the difficulties, this thesis tries to suggest possible approaches in developing more systematic way of protecting sensors based on the decomposition of each sensor attack into two types of basic factors, the beneficial property available from the sensors' measurand, and the unique fingerprint of physical signal emitters.