Secured cloud-based image data processing of self-driving vehicles using full homomorphic encryption

Self-driving vehicles leverage internet of things (IoT) technology to utilize multiple sensors to continuously monitor their environment and make decisions without human intervention. Data collected from these sensors require secure transmission and encrypted storage in cloud servers, where retrieving the data without decryption is necessary to ensure privacy. A novel method using full homomorphic encryption (FHE) image-based data is proposed, incorporating three different methods to search for the desired encrypted images stored in the cloud. This method involves converting images and objects detected by YOLOv9c into pixel data, then applying the Brakerski/Fan-Vercauteren (BFV) scheme to encrypt the data and store it on the cloud. Administrators can search through the encrypted images using the combination of techniques, including percentile similarity-based image detection, unknown object-based image detection, and known object-based image detection from encrypted images. FHE is used to provide various secure search approaches, contrasting with conventional index-based encrypted searching. The proposed solution provides a mechanism in the scenario of self-driving vehicles where object/image detection stored in the cloud can be done without decrypting it, hence enhancing privacy and security of data on the cloud generated by self-driving vehicles and IoT devices. • FHE-based approach enables secure image search from cloud. • YOLOv9c achieves high accuracy in object detection. • BFV scheme allows encrypted image retrieval without decryption. • Three search methods: similarity, known-object, and unknown-object detection. • FHE enables unlimited computations on encrypted images without decryption. • Supports secure and efficient image retrieval in cloud environments. • Evaluated for accuracy, efficiency, and resource consumption.