Leveraging Neuro-Inspired Reinforcement Learning for Secure Reputation-based Communication in Connected Vehicles

Secure communications in connected vehicles (CV) is essential to ensure the safety of drivers, passengers and pedestrians. As vehicles are becoming more connected and autonomous, they are reliant on communication and data exchange with infrastructure and other vehicles. Although Public Key Infrastructure has the potential to offer secure communication, it does not have ground truth information of vehicle location. Reputation-based communication can provide a more reliable approach to securing communication in a dynamic and constantly changing environment. This paper proposes a neuro-inspired reinforcement learning (RL) approach for reward estimation in CV networks. Vehicles estimate the reputation of neighboring vehicles by comparing broadcasted kinematic data with onboard sensor estimates along with connectivity topology inspired from brain, thereby forming a local graphical representation with reputation distribution. This information is shared with a centralized RL agent, which provides reward signals to each vehicle from combined reputation scores to incentivize accurate reputation estimates.