• autonomous driving
• cyber-physical systems
• cybersecurity
• digital twin
• neural networks

Cyber-Physical Systems (CPSs) are a large class of systems characterized by cooperating hardware and software components, connected with the external world.
Cybersecurity is a relevant activity in CPSs, since they are often safety-critical, and safety must be guaranteed also in case of cyber-attacks.
Modern autonomous vehicles are highly computerized CPSs, thus providing a wide range of access points for a potential attacker, who could gain full control over the vehicle and turn off all safety measures installed on it.
Our work provides a methodology designed to exploit a digital twin in co-simulation to gather data used for training and testing attack detection models based on machine learning algorithms. The developed approach has been applied to a case study having a digital twin composed of two vehicles, where the first one chases the second one. Results obtained are presented and analyzed, showing an accuracy of 94% in detecting attacks on the cyber-physical system, using a Multi-Layer Perceptron neural network.