A novel zero-watermarking algorithm based on multi-feature and DNA encryption for medical images

Different from the traditional watermarking schemes, zero-watermarking schemes are lossless embedding methods, which are applicable to be used in medical and other fields requiring high-integrity image copyright protection. However, with the improvement of telemedicine in the cloud, problems such as the leakage of patients’ personal information, the theft, and tamper have become increasingly prominent. How to more effectively ensure the information security and the copyright of medical images has attracted great attention of scholars in relevant fields. In this paper, combined with the characteristics of medical images, a robustness zero-watermarking scheme for medical images based on multi-feature and DNA encryption is proposed, focusing on how to improve its security and robustness. First, the medical image is divided into ROI (region of interest) and RONI (region of non-interest), and the ROI is further divided into layers. Then, CNN feature of ROI, the LBP, Gabor and variance features of each layer are extracted to enhance distinguishability and robustness. Moreover, a sine system based chaotic map is used to improve security of features and watermarks. Finally, DNA-based encryption, Zig-Zag and chaotic map are used to ensure the security of medical images. The proposed scheme can encrypt the features, watermarks and medical images effectively. The evaluation of the performance of encryption algorithms is compreshensive, including correlation analysis, which is close to 0 for the encrypted image; key sensitivity analysis, in which the similarity with slightly wrong key is below 0.5; information entropy, which is close to the theoretical value of 8 for encrypted images; encryption quality analysis with EQ of 380 and histogram analysis. The experimental results also show that this algorithm is highly robust to geometric attacks and common attacks, and most BER values and NC values under different attacks are about 0.01 and 0.99, respectively. It further highlights that the proposed scheme not only ensures the robustness and security, but also satisfies the lossless quality requirement, which outperforms the state-of-the-art schemes.