SMPP-CBIR: shorted and mixed aggregated image features for privacy-preserving content-based image retrieval

Abstract

Thanks to recent breakthroughs in photographic and digital technology, enormous amounts of image data are generated daily. Many content-based image retrieval (CBIR) systems have been developed for searching image collections. However, these systems need more computer and storage resources that can be met by cloud servers, since they supply a lot of processing power at a reasonable price. The protection of users' personal information is a worry for image owners since cloud services are not exactly trustworthy. In this paper, we suggest and put into practice a CBIR (SMPP-CBIR) technique for searching and retrieving ciphertext information that protects security. Asymmetric scalar-product-preserving encryption process (ASPE) is used to preserve aggregated mixed feature vectors while still enabling computation between them to describe the related picture collection. The k-means clustering algorithm is used to recursively arrange all encrypted attributes into a tree index in order to speed up search times. The findings show that SMPP-CBIR is more scalable, more precise, and faster in indexing and retrieval than earlier systems.