Privacy-Preserving Search over Encrypted Images

Abstract

The rapid development in the field of cloud computing, big data, and machine learning, motivates individuals and enterprises to outsource their multimedia and image data to the cloud. Although, outsourcing reduces the storage and computational overhead for resource constrained devices at client’s side, these services are still not getting attention due to security and privacy concerns. Clients care about the privacy of their data that is being shared and stored outside their jurisdiction. Fortunately, image processing in encrypted domain can overcome this issue. Current available techniques do not provide full privacy of image content, owner/client related information or have high computational cost. While retrieving the images from the cloud service provider (CSP), the client sends the query request to the CSP. These queries are not well protected and/or not randomized. Therefore, they are prone to traceability issues and do not provide security from search pattern leakage attacks. We propose a novel searchable encryption technique, which provides image content-based ranked searching and client privacy along with the un-tractability of client’s search queries. The scheme prevents against search pattern leakage attack as it is based on probabilistic trapdoors. Theoretical and experimental analysis shows that the proposed technique is more secure and efficient as compared to the state of the art schemes.