Thanks to its intuitive design, the picture hosting platform is quickly gaining users, although some worry about their privacy. Unfortunately, visual privacy protection solutions are typically permanent in their attempts to strike a compromise between privacy and usability. There is a lot of activity in the area of information security related to reversible data hiding (RDH). A cover medium can contain a secret in RDH. Applications that require deformation-free cover recovery in addition to hidden secret recovery highlight the importance of RDH over alternative data-hiding strategies. This paper presents a permutation ordered binary (POB) authentication and hiding scheme for medical images. The scheme consists of three steps: image recovering the original information with hidden information. Its purpose is to enhance the embedding ability of data in encrypted medical images and to protect medical image security. Two encrypted shares are created after compressing and encrypting the two new share pictures. With the use of the POB algorithm, secret data was embedded and authentication bits were attached to each of the two encrypted shares. During the recovery phase, the work presents a framework for ensemble models that utilises machine learning and deep learning to efficiently recover. Cover picture recovery is guaranteed by the majority vote of the various trained models. This study use the flamingo optimisation algorithm (FOA) to fine-tune the ensemble models' hyper-parameters. At last, the original medical image is restored after extracting the embedded hidden message. The results of the studies show that the strategy suggested in this study is superior when it comes to data embedding and recovery information. Excellent security has been achieved throughout the entire process, according to the experiments and analysis, using the proposed system. As data was being concealed, it managed to achieve high embedding capacity, PSNR, rate, and low SSIM