A high embedding efficiency RDH in encrypted image combining MSB prediction and matrix encoding for non-volatile memory-based cloud service

With the rise of non-volatile memory used in cloud service, it becomes urgent to reduce the write cost so as to improve the data manipulation capability. This paper proposed a high embedding efficiency reversible data hiding scheme for encrypted images (RDH-EI). It embeds data at the locations of most significant bit (MSB) using matrix encoding, resulting in a high embedding efficiency. In order to restore the image completely at the receiver side, a highly accurate MSB prediction based on orthogonal projection technique is preprocessed on the original image and the prediction result is self-embedded to help recover the image. The experimental results verify that the proposed scheme outperforms the previous related schemes, it can achieve a higher embedding capacity of 0.412 bits per pixel at average, moreover, a significantly higher embedding efficiency of 3.43 bits per change is obtained. That means the proposed scheme can reduce the write cost by minimizing the modification in the case of the same payload and is especially suitable for applications used the non-volatile memory. Meanwhile, the process of data extraction is error free and the image can be recovered completely.