Authenticating and securing healthcare records: A deep learning-based zero watermarking approach

Abstract

Security in medical records is critical to patient privacy and confidentiality. Digital Patient Records (DPR) hold sensitive information that can reveal a patient's health status and history. Their unauthorized access or exposure can lead to severe consequences, including identity theft, discrimination, and medical malpractice. Therefore, ensuring proper security measures is critical in protecting DPR and other medical records from breaches or unauthorized access. In this regard, a robust deep learning-based zero-watermarking approach is presented for authenticating and securing healthcare records. The carrier image is initially visibly marked with the hospital logo to identify ownership and prevent illegal duplication and forgery. The image mark is scrambled by applying the step space-filling curve method for improved security. In the final phase, Alexnet is used to extract the features of visibly marked carrier image. Further, NSST and SVD-based zero watermarking is implemented to conceal the scrambled mark within the features of visibly marked carrier images. It is essential for copyright protection since it establishes ownership while preventing the unauthorized use or dissemination of valuable medical research, images, and reports. The proposed framework has exhibited superior versatility, robustness, and imperceptibility compared to existing techniques with a maximum improvement of 47%.