Cognitive Risk Assessment Models for Cybersecurity in Autonomous Vehicle Operations

Abstract

Many autonomous vehicle (AV) functions have already been defined and have demonstrated a positive potential impact if properly adapted, including platooning and cooperative adaptive cruise control (CACC) techniques [1]. Functional, reactive and object-oriented (FRaOOP) requirements, like primary safety (PS) and physical safety (PhS) and cyber safety (CS), have been taken into account in the cybersecurity literature, thus far excluding the characterization of human intelligence which is a key requirement as well. The current cybersecurity validation test benches combine the Reinforcement Learning approach to simulate an adversarial strategy for their implementation and a framework with the Fault Injection approach. Also, in the Authors’ literature an automatic evaluation methodology for the test bench was investigated with the goal to assess countermeasure effectiveness. This paper enhances the research presented in the literature with a novel and innovative approach referred to as the cognitive cybersecurity framework, where the human intelligence characteristic has been introduced in the validation test bench. This approach allows the AV companies to evaluate the criticality of the residually exploitable cybersecurity weakness, thus improving the design-phase criticality mitigation and countermeasures ranking process by a V2I cognitive reticularity methodology.