Optimized Deep Learning-Enabled Hybrid Logistic Piece-Wise Chaotic Map for Secured Medical Data Storage System

Background: In recent times, medical technology has generated massive reports such as scanned medical images and electronic patient accounts. These reports are necessary to be stored in the highly secured platform for further reference. Traditional storage systems are infeasible for storing massive data. In addition, it suffers to provide secure storage and privacy protection at the time of medical services. It is necessary to provide secure storage and full utilization of personal medical records for the common people in practice. The healthcare system based on IoT enhances the support for the patients and doctors in diagnosing the sufferers at an accurate time using the monitored health data. Yet, doctors make an inappropriate decision regarding the sufferer’s sickness when the information regarding health data saved in the cloud gets lost or hacked owing to an external attack or also power failure. Hence, it is highly essential for verifying the truthfulness of the sufferer’s information regarding health data saved on the cloud.Hypothesis: The major intention of this task is to adopt a new chaotic-based healthcare medical data storage system for storing medical data (medical images) with high protection. Methodology: Initially, the input medical images are gathered from the benchmark datasets concerning different modalities. The collected medical images are enciphered by developing Hybrid Chaotic Map by adapting the 2D-Logistic Chaotic Map (2DLCM), and Piece-Wise Linear Chaotic Map (PWLCM) referred to as Hybrid Logistic Piece-Wise Chaotic Map (HLPWCM). An Optimized Recurrent Neural Network (O-RNN) is proposed for key generation using Best Fitness-based Coefficient vector improved Spotted Hyena Optimizer (BF-CSHO). The O-RNN-based key generation utilizes the extracted image features like first and second-order statistical features and the targets are acquired as a unique encrypted key, which is used for securing the medical data. The same BF-CSHO is used for improving the training algorithm (weight optimization) of RNN to minimize the Mean Absolute Error (MAE) between the cipher (encrypted) images and original images. Results: From the result analysis, the suggested BF-CSHO-RNN-HLPWCM, by considering the image size at 554×554 shows 10.4%, 8.5%, 3.97%, 0.62%, 3.88%, 2.40%, and 7.82% provides better computational efficiency than LCM, PWLCM, LPWCM, PSO-RNN-HLPWCM, JA-RNN-HLPWCM, GWO-RNN-HLPWCM, and SHO-RNN-HLPWCM, respectively. Conclusion: Thus, the simulation findings show the effective efficiency of the offered method owing to the security of the stored medical data.