Evaluation of a Reputation Management Technique for Autonomous Vehicles

Abstract

Future autonomous vehicles will rely heavily on sharing and communicating information with other vehicles to maximize their efficiency. These interactions, which will likely include details about the positions of surrounding vehicles and obstacles on the road, are essential to their decision-making and the prevention of accidents. However, malicious vehicles—those that intentionally communicate false information—have the capacity to adversely influence other vehicles in the network. This paper presents and evaluates a reputation management system, capable of identifying malicious actors, to mitigate their effects on the vehicle network. The viability of multiple report weighting schemes to calculate reputation is evaluated through a simulation, and a blockchain-based backend for the reputation management system to securely maintain and communicate reputation data is proposed. Storage and computational challenges are considered. This paper shows that weighting schemas, related to the number and reputation of witnesses, positively affect the accuracy of the model and are able to identify malicious vehicles in a network with consistent accuracy and scalability.