Fuzzy and Spiking Neural Network-based Secure Data Exchange Framework for Autonomous Vehicle

Autonomous vehicles (AVs) represent a revolutionary step in intelligent transportation (ITS), which uses artificial intelligence (AI) models to navigate without human intervention. However, the reliance on sensor networks for operational functionality introduces significant vulnerabilities as inadequate training of these sensors can lead to sub-optimal path selection or, more critically, accidents. The main security issue addressed in this study is the lack of strong security mechanisms for the operation of sensors in AVs, which makes them vulnerable to counter-revolutionary actions. These attacks can degrade performance and compromise the security of ITS. Thus, we propose a novel intrusion detection system integrated with fuzzy logic and AI, which is designed to validate and sanitize the communication between the sensors to improve the performance of AVs. For that, a sensor communication dataset is utilized comprised of malicious and non-malicious data. Furthermore, a spiking neural network (SNN) is employed to get trained on the dataset for efficient binary classification tasks. Our implementation of the SNN model has demonstrated a remarkable accuracy of 97.76% in identifying non-malicious data. The performance of our framework is further validated by comprehensive evaluations using indicators such as recall, accuracy, precision, and F1 score.