Machine Learning Approach for Securing Autonomous Vehicle Applications in Urban Networks

In recent years, Vehicle-to-Vehicle (V2V) communication has gained significant traction in the automotive industry due to its ability to monitor key metrics such as vehicle speed, proximity to nearby vehicles, and road conditions in real time. This technology greatly enhances road safety and traffic management by enabling continuous information exchange between vehicles. However, conventional systems often face challenges when monitoring road safety and traffic flow in dense urban environments. To overcome these limitations, this study proposes a novel autonomous vehicle application security framework that integrates quantum cryptography with cloud-based networks, supported by noise deciphering through a machine learning approach. The framework draws on three types of input data sources, each containing distinct attack types, to monitor and analyze the autonomous vehicle network for malicious activity. Authentication is carried out using a convolutional federated quantum key encryption algorithm within the cloud environment. To handle cloud network noise, the system employs a fast Fourier Gaussian encoder model. By combining historical and real-time behavioral data, the system evaluates the reliability of traffic data and network nodes, enabling accurate anomaly detection. Experimental evaluations were conducted using multiple autonomous vehicle security datasets, assessing metrics such as detection accuracy, noise ratio, average precision, Quality of Service (QoS), and data integrity. The proposed method achieved 90% average precision, 89% data integrity, 95% QoS, 96% detection accuracy, and a 65% noise ratio. These findings demonstrate that the proposed framework delivers enhanced security for V2V applications compared to existing solutions.