Anomaly Detection Framework for Securing Next Generation Networks of Platoons of Autonomous Vehicles in a Vehicle-to-Everything System

Abstract

We consider a security setting involving a platoon of autonomous vehicles (AVs) that commute from one place to another. Such vehicle platooning is utilized to optimize the usage and safety of highways. We propose a dynamic framework for a network of platoons that captures both the communication between different platoons along with the communication between different AVs within the single platoon. We propose an authenticity score scheme for monitoring the behavior of the platoons. We also propose a two-phase anomaly detection within a single platoon to elect and maintain a benign platoon leader. We then propose a long-short term memory (LSTM)-based RSU level anomaly detection scheme to safeguard the whole network of platoons. Finally, we adapt group-based signatures and channel switching schemes for ensuring that the communication channels between AVs and platoons stay secure against man-in-the-middle and denial of service attacks. We perform extensive numerical simulations to evaluate the different components in our framework.