MSPPIR: Multi-Source Privacy-Preserving Image Retrieval in cloud computing

Abstract

Content-Based Image Retrieval (CBIR) techniques have been widely researched and in service with cloud computing like Google Images. However, the images always contain rich sensitive information. In this case, privacy protection becomes a big problem as the cloud always cannot be fully trusted. Many privacy-preserving image retrieval schemes have been proposed, in which the image owner can upload the encrypted images to the cloud, and the owner himself or the authorized user can execute the secure retrieval with the help of the cloud. Nevertheless, few existing researchers notice the multi-source scene that is more practical. In this paper, we analyze the difficulties in Multi-Source Privacy-Preserving Image Retrieval (MSPPIR). Then we use the image in JPEG-format as the example, to propose a scheme called JES-MSIR, namely a novel JPEG image Encryption Scheme which is made for Multi-Source content-based Image Retrieval. JES-MSIR can support the requirements of MSPPIR, including the constant-rounds secure retrieval from multiple sources and the union of multiple sources for better retrieval services. Experiment results and security analysis on the proposed scheme show its efficiency, security, and accuracy.