Security analysis of a reversible watermarking algorithm for encrypted images in wavelet domain

Abstract

In recent years, reversible data hiding in encrypted images (RDH-EI) has attracted more and more attention, and many RDH-EI algorithms have been proposed. Scholars focus on improving embedded capacity and decreasing bit error rate, but the security issue of algorithms has not been paid enough attention to. However, the security of algorithms is a core problem. Recently, a RDH-EI algorithm based on compressed sensing and discrete wavelet transform is proposed. And considering the watermark embedding position keystream is related to the plain-image, the proposed algorithm is supposed to have high security. However, we find its security flaw, and by using chosen-plaintext attacks and differential cryptanalysis, we can extract all the encryption keys and all the embedding keys with merely 14 spatial images. In particular, compared with other methods, the number of images required in our method is independent of image size, which means that it requires less cost to break the algorithm. Finally, to resist the proposed attack, two improved suggestions are put forward. Simulation results demonstrate that the security of the improved algorithm is enhanced.