Privacy Preserving Data Sharing for Connected Autonomous Vehicles Leveraging Edge Computing
Tianyu Bai · 2025
Connected Autonomous Vehicles (CAVs) integrate autonomous driving with advanced connectivity, enabling real-time communication with other vehicles, infrastructure, and cloud systems. This connectivity enhances safety, efficiency, and navigation but also introduces significant privacy challenges. CAVs rely on environment perception, using sensors such as cameras, LiDAR, radar, and GPS to gather data for navigation and decision-making. While edge computing allows CAVs to offload computationally intensive perception tasks to edge servers, it raises concerns about data security, as sensitive sensor information is transmitted and processed externally. This dissertation presents a privacy-preserving framework for CAV environment perception and data sharing, ensuring that sensitive information remains protected during processing and transmission. The proposed approach enables secure object detection, data sharing, and deep learning inference without exposing raw sensor data. A privacy-preserving perception framework allows CAVs to perform image-based object detection securely while leveraging the computational power of edge servers. A privacy-enhancing data-sharing mechanism enables users to define and remove sensitive image and video content before transmission. Additionally, privacy-preserving deep learning models, including a secure driver monitoring system and a privacy-preserving large language model, ensure that sensitive driver and user data remain protected during inference. The dissertation also addresses the trade-off between privacy and computational efficiency. While secure computation protocols introduce additional latency, the proposed frameworks are designed to minimize overhead while maintaining high accuracy in perception and decision-making. The research contributes to the advancement of privacy-preserving technologies in CAVs, providing scalable and efficient solutions that balance security with real-time processing requirements.