Multi-PVO for Reversible Data Hiding Based onFull Homomorphic Encryption

Abstract To enhance the embedding performance of reversible data hiding algorithms and bolster the security of data transmission processes, a novel pixel value ordering method based on full homomorphic encryption is proposed in this paper. First, we propose a multi-PVO method to improve the fidelity of images with equal embedding capacity. In addition, we propose a double layer encryption method and achieve Paillier's fully homomorphic encryption. In practice, the sender encrypts the original image and then transmits the ciphertext to the embedding side. The embedding side performs a multi-PVO method and data hiding based on homomorphism to seamlessly embed the data into the ciphertext. This implementation offers dual functionality. On the one hand, users with no private keys can directly embed or extract the secret data without decryption keys. On the other hand, users with private keys can decrypt the marked ciphertext and help retrieve the marked plaintext. Finally, utilizing multi-PVO for extraction or recovery, additional data or original image can be retrieved from the marked image at receiver. With 20,000 bits embedded, the PSNR of the image Lena is as high as 59.60 dB, which demonstrates great image quality with the same embedding capability.