A Robust Zero-Watermarking Algorithm for Encrypted Medical Images in the DWT-DFT Encrypted Domain

Abstract

In order to protect personal information, numerous works has been done in watermarking field. However, there still leaves some problems to be solved: (1) most of the watermarking methods were processed in the plaintext domains, which leave latent risk of exposing host image information, thus it is needed to encrypt the host image and process the watermarking scheme in the encrypted domain; (2) numerous image encryption methods had been searched, while not all of them can meet the robustness requirements when applied in the encrypted domain; (3) for some special fields of watermarking applications, medical images, for example, image integrity is an important criterion that should be strictly taken into account. Thus, that kind of watermarking methods which applies by modifying the pixel values are not suitable in this situation. In order to achieve information hiding in such kind of images, special techniques which do not change image integrity is needed. (4) By utilizing homomorphic encryption scheme, one can process watermark extraction without decrypting the encrypted watermarked image first, while it cost too much time in image encryption and decryption, the computational speed need to be improved. Based on the points mentioned above, we proposed a robust zero-watermarking scheme in the DWT-DFT encrypted domain, which embeds and extracts watermark without modifying the pixel values. Firstly, we encrypted both original medical image and watermark image. Then, we extract the DWT-DFT low frequency coefficients as encrypted medical images’ feature vector. In watermark embedding and extraction phases, we adopt zero-watermarking technique to ensure integrity of medical images. Taking “db2” wavelet transform for example, we conduct the experiments on the visual quality and robustness of our watermarking scheme. Experimental results demonstrate that our algorithm achieves not only good watermarking robustness, but also ideal computation speed in the homomorphic encrypted domain.