Efficient and secure medical big data management system using optimal map-reduce framework and deep learning

Efficient and secure medical big data management system using optimal map-reduce framework and deep learning

연구 분야: Cryptography

학회: Multimedia Tools and Applications

Big Data (BD) and cloud computing (CC) are the two widely used technologies and focus of study in several industries. Amongst all, healthcare sources generate a tremendous amount of data daily. Traditional processing techniques cannot handle this data because they are huge. Furthermore, being large, this data is also dynamic and diverse. Large data sets are stored, processed, and analyzed under BD. However, as the volume of data increased, third parties could hack it easily. The data are regularly stored in the cloud in an encrypted form to protect the data from intruders. This paper proposes a secure and efficient medical BD management and classification scheme using optimal Map Reduce (MR) and deep learning framework in a cloud environment. The system comprises '4' phases: authentication of patients, BD management in the cloud, secure data transfer and BD classification. First, the patient who wants to upload the files to the cloud and access the resources from the cloud is registered with the TC. It generates hash value using the Whirlpool Hashing (WH) algorithm, and the user credentials are stored in Blockchain (BC) to protect the network from unauthorized access. Once the authentication is successful, the patient can access any data in the cloud. Before uploading the file to the cloud, the preprocessing is carried out using missing values imputation, numerical conversion, and normalization, which improves the data quality. It is fed into the MR framework using Kernelized K-Means (KKM) clustering and Enhanced Butterfly Optimization Algorithm (EBOA) for managing the data efficiently. Then the map-reduced data is encrypted using Whirlpool Hashing-based Enhanced Rivest Shamir Adelman (WHERSA) to upload into the cloud securely. Finally, the classification of BD is done using Enhanced Ant Colony Weight Optimization based Deep Belief Network (EACWODBN) for disease prediction. The experimental outcomes reveal that the proposed system outperformed state-of-art methods while offering effectual and secure data management.