Detecting Dark Intelligence in Cyber‐Physical Digital Twins Using Self‐Learning AI and Cognitive Security Analytics

This chapter analyses self-learning cyber-physical digital twins (CPDTs) as innovative systems which enable the identification and resolution of dark intelligence threats targeting essential systems such as smart grids and autonomous vehicles and industrial internet of things networks. It proposes that universities establish multimodal federated reinforcement learning systems and quantum-safe analytics tools and develop healthcare twins which will enable CPDTs to function as partners who protect cognitive resilience through their technical capacity and their ethical governance systems which safeguard cyber-physical ecosystems. In stealth attack detection for CPDTs, there is an emphasis on detecting adversaries that attempt to minimize their detectability while still controlling the cyber-physical system to unsafe states. Model poisoning in CPDTs attacks the learning and decision-making processes of the DT, with the aim of corrupting the models while still appearing to perform well on benign data.