Bio-inspired Active System Identification: a Cyber-Physical Intelligence Attack in Networked Control Systems

From the point of view of the control theory, the literature indicates that stealthy and accurate cyber-physical attacks on Networked Control System (NCS) must be planned based on an accurate knowledge about the model of the attacked system. However, most literature about these attacks does not indicate how such knowledge is obtained by the attacker. So, to fill this hiatus, an Active System Identification attack is proposed in this paper, where the attacker injects data on the NCS to learn about its model. The attack is implemented based on two bio-inspired metaheuristics: Backtracking Search Optimization Algorithm (BSA) and Particle Swarm Optimization (PSO). To improve the accuracy of the estimated models, a statistical refinement is proposed for the outcomes of the two optimization algorithms. Additionally, a set of data injection attacks are shown in order to demonstrate the capability of the proposed attack in supporting the design of other sophisticated attacks. The results indicate a better performance of the BSA-based attacks, especially when the captured signals contain white Gaussian noise. The goal of this paper is to demonstrate the degree of accuracy that this System Identification attack may achieve, highlighting the potential impacts and encouraging the research of possible countermeasures.