Fidelity Preserved Data Hiding in Encrypted Highly Autocorrelated Data Based on Homomorphism and Compressive Sensing

Abstract

Images and videos (consisting of successive images) are highly autocorrelated due to the inherent spatial correlations. However, when the data are encrypted for privacy protection, the spatial correlations would be eliminated. Thus, it is difficult for an untrusted third party in the cloud to embed annotation information or auxiliary information in the encrypted media. In this paper, we proposed a novel method to maintain the spatial correlation in the encrypted domain to some extent by using a homomorphic cryptosystem in order to achieve high-quality data hiding. The textured and smooth blocks of the image can be identified in the encrypted domain. With a compressive sensing technology, the LSB layers of smooth blocks can be well-handled to make room for accommodating additional data. In the data hiding process, data hider embeds the additional data into LSBs of the smooth area to preserve high fidelity of the stego-image. On the receiver side, the embedded data can be extracted without any distortion from the encrypted image only by data-embedding key, and also, we can reap the directly decrypted image with a high visual quality only with the encryption key. In the case of the user possesses both the encryption key and the data-embedding key, the additional data can be extracted accurately, and the image recovery with overwhelming probability can be achieved. The vast experimental results manifest that the proposed method not merely has excellent security performance but also maintains the fidelity of the host image while providing considerable embedding capacity, which is superior to other state-of-the-art schemes.