Abstract
Future transportation systems are expected to be transformed by connected vehicles, offering greatly improvements in traffic efficiency, mobility, and road safety. However, realizing these benefits requires not only seamless connectivity but also robust security to ensure there liable exchange of extensive sensor data and contextual information among vehicles, infrastructures, and surrounding environments. To this end, the integration of advanced communication and sensing technologies is essential for paving the way toward next-generation autonomous mobility. In this dissertation, four key contributions are presented. First, it provides a comprehensive performance assessment of mmWave-enabled vehicle-to-vehicle communications, evaluating achievable throughput, latency, and reliability metrics. Second, a deep learning based solution is presented so that the optimal beams having sufficient mmWave received powers can be estimated by utilizing multi-modality sensing data. Third, a transformer-based framework is introduced, where multi-head cross-modal attention is utilized to capture dependencies and correlations between different sensing modalities, and subsequently fused the multimodal features to improve beam prediction. Finally, it develops an attack-defense tree methodology to systematically evaluate cyber security vulnerabilities in connected vehicles. Together, these contributions aim to advance the design of safe, intelligent, and resilient transportation systems by bridging reliable connectivity with strengthened security mechanisms for connected vehicles.