Federated Learning Approaches for Collaborative Threat Detection in Autonomous Vehicle Networks

Abstract

The robustness of the AV should be guaranteed at multiple levels: (i) at the isolated level of a particular AV in response to cryptic communications from sensors and actuators and against stealthy attacks of methods; (ii) at the level of all distributed components (collaborative) of the AV fleet (including the remote operator or AV operational services) against threats that are capable of affecting the global functionality of the system of the AVs. It is clear that the security and safety of the AV fleet becomes crucial in the event of the verification of infrastructure-grade services (including human safety and private data security). The deployment and efficient functioning of the installations of the AV fleet are likely to occur through the establishment of future cooperative testbeds for safety assessment aimed at creating a completely acceptable technology by the corresponding standards (e.g., necessary conditions of the ISO 26262:2018 and the language of the SOTIF standard: preliminary work of the ISO/PAS 21448).