Reversible data hiding in homomorphically encrypted image using interpolation technique

In this paper, a reversible data hiding scheme based on interpolation technique for encrypted images by using homomorphic and probabilistic properties of Paillier cryptosystem is presented. At first, the image owner generates a location map by using an interpolation technique to estimate the Most Significant Bits (MSBs) of pixel to find whether a pixel can be used for embedding or not. Next, with the help of the location map, the original image is preprocessed to create some spare space for data embedding. Meanwhile, the location map is compressed losslessly and the information of compressed location map is substituted with Least Significant Bits (LSBs) of border pixels. Furthermore, the preprocessed image is encrypted by Paillier cryptosystem and sent to the data hider along with location map and original LSBs of border pixels. At the data hiding phase (note that original image cannot be accessed), the additional data and LSBs of border pixels are embedded into homomorphic encrypted image using location map. At the receiver side, the embedded additional data and original image are recovered losslessly in non-separable manner (i.e., from directly decrypted image). Experimental results demonstrate feasibility and efficiency of the proposed scheme, particularly in embedding, image recovery and performance on different security key size.