A novel scheme to encrypting autonomous driving scene point clouds based on optical chaos

Abstract

The widespread adoption of 3D point cloud technology in autonomous driving has raised concerns about the potential leakage of private information among Internet of Vehicles (IoV) users, especially when data is exchanged between vehicles without adequate protection. This paper introduces a novel encryption and decryption scheme for 3D point cloud data, designed to address security and privacy concerns in autonomous driving environments. The optical system, based on vertical-cavity surface-emitting lasers (VCSELs), is configured to generate optical chaos, which is then applied to the permutation and diffusion of 3D point clouds. In the case study, 3D point cloud images from the KITTI dataset are encrypted and decrypted, and the three classes of objects — cars, cyclists, and pedestrians — are detected in the original, encrypted, and decrypted datasets using the Point-Voxel Region Convolutional Neural Network (PV-RCNN). The mean average precision (mAP) for the encrypted dataset is nearly zero, indicating that the 3D point cloud objects cannot be detected. In contrast, the mAP for the decrypted dataset closely matches that of the original dataset, demonstrating the effectiveness and feasibility of the proposed privacy protection scheme. Additionally, a detailed security analysis of the geometric features in 3D point clouds confirms that the scheme provides robust security and privacy protection for the scene information in 3D point cloud images.