Adaptive security reinforcement learning for enhanced cryptographic vehicle-to-vehicle communication security

Secure and reliable Vehicle-to-Vehicle (V2V) communications are critical in the age of connected vehicles and intelligent transport systems. This paper introduces a new technique, Adaptive Security Reinforcement Learning (ASRL), which aims at adjusting and learning cryptographic techniques according to effective on-the-fly information about vehicles. Marine security settings can be automatically adjusted via ASRL by using reinforcement learning to optimize set security settings, such as Elapsed Time, Speed, RPM, Acceleration, Engine Load, Fuel Level and Temperature, given changing conditions. The purpose of this adaptive approach is to keep safety and efficiency in balance, so that the overall security of vehicular communication can be improved without reducing operational performance. The solution involves multivariate polynomial lattices (MPL) for strong cryptographic underpinnings, quantum resilient cryptography for secure models against yet-to-become quantum-based threats, and discretized mathematical implies for optimized algorithms. The system also uses state-of-the art techniques to withstand algebraic and combinatorial attacks, including the use of designs and Boolean functions. The paper presents the conceptual framework of ASRL, by defining state space, action space and reward function in the idealized scenario and describing learning algorithm in detail, together with simulations to illustrate its effectiveness. Through this holistic treatment, we provide a working solution to evolving security threats in V2V communications, and hence we enable safer and more resilient vehicular networks.