FLPPBC: A Federated Learning-Driven Privacy-Preserving Model Using Private Permissioned Blockchain for Healthcare Systems

Objectives: To utilize federated learning and blockchain for developing a layered security model to offer protection and discretion of patient data in a private healthcare system. Method: To improve the security and confidentiality in smart healthcare, this work proposes a Federated Learning-driven Privacy Preserving Blockchain (FLPPBC) model which contains three layers: data, blockchain and aggregated. In data layer, the homomorphic encryption algorithm is used to encrypt the data and local model training parameters. The gradient boosting algorithm is used to generate the local training model. The SHA-256 hashing method and Proof of Authority (PoA) consensus mechanism is employed in blockchain layer to provide the safety and privacy. The weighted average aggregation is utilized in aggregated layer to build the global model. The medical dataset heart disease and breast cancer from UCI repository is utilized to evaluate the effectiveness of FLPPBC in terms of accuracy, precision, recall, f-measure, latency and throughput. Findings: FLPPBC utilized healthcare datasets (heart disease and breast cancer) to analyze the performance. It attains more than 98% accuracy for health data. FLPPBC also achieves minimum latency (39.27 ms) and maximum throughput (125 Trans/Sec). The approach efficiently handles many transactions and adds them to the blockchain quickly, reducing latency. Novelty: FL efficiently trains decentralized data using machine learning models, and homomorphic and BC enables secure access to distributed patient data without compromising confidentiality. The proposed approach uses gradient boosting for healthcare data classification, and SHA-256 and PoA provide the security model. Keywords: Blockchain, Federated learning, Healthcare, Homomorphic encryption, Privacy-preserving, Data security