Noise-free thumbnail-preserving image encryption based on MSB prediction

Abstract

People are increasingly storing their images on the cloud as cloud storage grows in popularity, delivering a variety of benefits. However, the associated privacy concern may pose a serious challenge to data users. While traditional confusion-diffusion-based encryption is an effective way to resolve this concern by yielding a noise-like image without visual meaning, it sacrifices visual usability, i.e., it fails to make privacy and usability compatible. Thumbnail-preserving encryption (TPE) is proposed to balance privacy and usability, which requires that the thumbnail of the encrypted image is identical or approximate to that of the original. Although approximate-TPE has higher efficiency than ideal-TPE, the existing approximate-TPE methods cannot decrypt without loss. Motivated by this, we design a novel approximate-TPE method that is noise-free for the first time, which employs reversible data hiding based on the most significant bit prediction. Experimental results show that the method can effectively eliminate noise and finally perform decryption without sacrificing the quality of the decrypted image. Furthermore, image owners can recognize their images through the preserved coarse perceptual characteristics in the encrypted image, though it is difficult for an unauthorized third party to successfully discern with limited prior knowledge.