MultiMark: Multi-Descriptor Fuzzy Assisted Secure NIfTI Image Transfer Framework With Features Control Authentication

The communication technology has recently advanced rapidly. This advancement aims to provide secure transmission with better reliability, especially in telemedicine applications, where secure and reliable transmission is necessary. This article presents a novel dual image watermarking technique for the integrity protection of medical records. This technique employs a redundant multiresolution domain for the watermarking process. Different sub-bands are chosen after the entropy computation of the host image. Efficient and more stable lower–upper (LU) decomposition is applied successively over the previously obtained sub-bands to choose a more stable matrix. Two hybrid watermarks are generated using the patient identity and the medical logo for embedding. A multi-descriptor fuzzy inference system (FIS) is used to compute the optimal scaling factor. Texture, entropy, and change in image per pixel (CIPP) are selected as the membership functions for FIS. The performance of the proposed method is verified against a different set of attacks. Robustness is also verified with a denoising convolutional neural network (DnCNN) for the presented method. The average improvement in robustness is 25.50%, while 35.75% in imperceptibility. In the proposed method, KAZE features are also successfully matched for effective and efficient authentication.