Blockchain-Based Federated Learning-Convolutional Neural Network for Preserving Data Privacy and Security

The Internet of Things (IoT) and Industrial cloud computing have completely transferred the healthcare sector due to the rapid rise of distributed healthcare data. The security and privacy of healthcare data are critical issues facing in the healthcare sector. In this research, a BlockChain-based Federated Learning-Convolutional Neural Network (BC-FL-CNN) is proposed for preserving electronic health data privacy by integrating the BC and Deep Learning (DL) approach. CNN is used to classify normal and abnormal users in the processed dataset. The abnormal users were then processed and removed from the database as well as access to the health data utilizing BC with FL approach. Existing methods such as Convolution Neural Network-BlockChain-Cryptography-Federated Learning (CNN-BC-Cryp-FL), FL, and FL Risk-based Authorization Middleware for Healthcare (FRAMH) are used to compare with the proposed BC-FL-CNN approach. The proposed BC-FL-CNN achieves better accuracy of 98.25%, precision of 97.23%, recall of 97.56%, and 97.16% f1-score when compared to the existing approaches like CNN-BC-Cryp-FL, FL, and FRAMH respectively.