Contextual Based Attack Detection and Resiliency Policies for Autonomous Vehicles Platoons

The goal of this work is to implement resilience policies for Autonomous Vehicles under attack, based on the context in which the platoon is located. In a platoon model, we define the leader as the fulcrum entity of the whole model, which acts as a guide for the other entities, defined as followers. To allow the vehicles to communicate with each other and exchange the data necessary to establish an autonomous driving scenario, a system of data transmission from and between the entities is implemented. Considering this transmission as unsafe and attackable from the outside, the implemented policies recognize a possible attack and define various rules so that the targeted subject can recover and continue working in a safe driving scenario. The driving contexts are implemented thanks to the Carla simulator and collect the various nuances and meanings of real driving, allowing us to define realistic attacks and scenarios and to observe and analyze the various responses. Each context is therefore repeatedly tested and analyzed, both in situations of regular operation and in the presence of various attacks, to ascertain the functioning and accuracy of the rules above. In conclusion, a report on the mentioned tests is indicated, reporting the precision of the rules and their practical functioning, supported by the data collected during the simulations.