Separable Reversible Data Hiding in Encrypted Image with Classification Permutation

To improve the security and the quality of decrypted and recovered images, this work proposes a classification permutation based separable reversible data hiding in encrypted image. A classification permutation encryption combining with the XOR-encryption is firstly designed to improve the privacy of both pixel-value and pixel-location. And then, each bit in the additional data is embedded in the most significant bit (MSB) of the encrypted pixels which is randomly chosen in the smooth set based on the data-hiding key. Since the MSB of smooth pixels are used to embed the additional data, the quality of decrypted and recovered image can be significantly improved by exploiting the spatial correlation. Experimental results demonstrate that the quality of decrypted image obtained by the proposed method outperforms that of the existing methods, and the probability for lossless recovery of proposed method is 100% even if for texture image Baboon at a high embedding rate of 15%.