The main features of Cyber Physical Systems (CPS) are the joint use of hardware and software resources for computing, communication and control. Among the several applications of CPS, we can identify the following systems: high reliability medical devices and systems, traffic control and safety, advanced automotive systems, industrial process control, environmental control, avionics, critical infrastructure control, distributed robotics, defence, etc. In this kind of systems, due to the tight interdependence of cyber-physical coupling, security is a major concern. Indeed, with the increased complexity of attacks, it is necessary for defenders to be able not only to detect malicious activities when they occur, but also to predict the specific actions that will be taken by an adversary during an attack. The GNADIS project aims at investigating new AI techniques such as GNNs (Graph Neural Networks) to both detect attacks and identify attack paths that naturally have a graph structure so that to predict, by extrapolation, potential attacker targets.
