Method for vehicle-theft driver's identification with privacy protection

Abstract

This invention pertains to the field of vehicle anti-theft technology and discloses a vehicle anti-theft method based on encrypted facial recognition. The method includes: periodically capturing facial data of the driver within the vehicle, using an in-vehicle neural network-based facial feature extraction model to extract feature vectors, and comparing these vectors with a preset whitelist to confirm authorization. If the driver is an unauthorized person, the feature vector is transmitted to the internet of vehicles for detection, where abnormal recognition triggers an alarm and uploads the data to a trusted authority within the internet of vehicles for tracking. The internet of vehicles consists of a trusted authority, auxiliary servers, cloud servers, roadside units, and vehicles, all of which deploy physically unclonable functions. This system enables rapid identity verification and anomaly detection to ensure driving safety. It can resist side-channel attacks on physical devices, capture attacks, and collusion attacks on servers, thereby protecting equipment safety and data privacy.