Application of probabilistic inference to resiliency and security analysis of cyber-physical systems

Abstract

This dissertation studies two important topics regarding the resiliency and the security of cyber-physical systems (CPSs). In the first work, a self-healing graphical representation is proposed to study the contagion of failures in self-healing interdependent networks. To this end, a graphical model representation of an interdependent cyber-physical system is proposed, in which nodes denote various cyber or physical functionalities, and edges capture the interactions between nodes. Then, a message-passing (belief propagation) algorithm is applied to this representation in order to analyze network reactions to initial disruptions. The framework is then extended to cases where the propagation of failures in the physical network is faster than the healing responses of the cyber network. Such scenarios are of interest in many real-life applications, such as the smart grid. As a result, it is proven that as the number of message-passing iterations increases, the network reaches a steady-state condition that would be either a complete healing or a complete collapse. The findings from this analysis help network designers have a better understanding of the resiliency of CPSs. In the second work, security measurement and the malicious node detection of autonomous vehicles in intelligent transportation systems are studied. First, a simple security model based on Bayesian defense graphs is proposed to quantitatively assess the likelihood of threats against autonomous vehicles (AVs) in the presence of available countermeasures. Then, a game-theoretic model is represented using a local voting-based game to detect misbehaving neighboring vehicles in places where centrally managed stations are absent. In order to capture the inherent uncertainty of vehicles in ephemeral vehicular networks, a Bayesian game is used in which malicious nodes can potentially impact the result of the game. Then, equilibria of this game are obtained to study the strategies of malicious and benign nodes in networks. Using the analysis, the game parameters can be designed to achieve the maximum performance of misbehavior detection in vehicular networks.