A coalitional security game against data integrity attacks in autonomous vehicle networks

Abstract

Autonomous Vehicles (AV) possess an automated driving system that allows the vehicle to respond to external conditions and provide advanced and intelligent transportation services with limited to no human intervention. For AV, the legitimacy and reliability of sensed and transmitted data have direct influence on road network operations. For instance, the integrity of vehicle communications is vulnerable to various threats such as spoofing attacks. Due to the extremely volatile nature of AV, traditional centralized security solutions lack in flexibility and mobility support. Moreover, the very large scale of future AV networks mandates an innovative design of intelligent communications among the vehicles to identify authentic traffic information broadcasted all across the network. In this paper, we design a vehicular collaboration framework that drives the interactions among autonomous vehicles based on trustworthy communications to prevent data integrity attacks. The design involves a distributed, hedonic coalition formation game in which the vehicles evaluate their utilities according to their trust in each other and the data they transmit. The proposed vehicular coalition formation algorithm enables the vehicles to dynamically join and leave the collaborative communities based on trust updates. Vehicles' hybrid trust relationships are derived by integrating the three parameters of the REK model: Reputation, Experience, and Knowledge to optimally exploit the available information in the network and adapt the framework to different road traffic contexts. Furthermore, to protect the formed coalitions against data integrity attacks, a trust-based data aggregation procedure using Dempster-Shafer theory of evidence is adopted to limit the effect of false events on vehicles' decisions. The experimental results show that our collaboration framework is resilient to data corruption and fabrication attacks.