Towards Efficient and Trustworthy Pandemic Diagnosis in Smart Cities: A Blockchain-Based Federated Learning Approach

Abstract

In the aftermath of the COVID-19 pandemic, the need for efficient and reliable disease diagnosis in smart cities has become increasingly serious. In this study, we introduce a novel blockchain-based federated learning framework tailored specifically for the diagnosis of pandemic diseases in smart cities, called BFLPD, with a focus on COVID-19 as a case study. The proposed BFLPD takes advantage of the decentralized nature of blockchain technology to design collaborative intelligence for automated diagnosis without violating trustworthiness metrics, such as privacy, security, and data sharing, which are encountered in healthcare systems of smart cities. Cheon–Kim–Kim–Song (CKKS) encryption is intelligently redesigned in BFLPD to ensure the secure sharing of learning updates during the training process. The proposed BFLPD presents a decentralized secure aggregation method that safeguards the integrity of the global model against adversarial attacks, thereby improving the overall efficiency and trustworthiness of our system. Extensive experiments and evaluations using a case study of COVID-19 ultrasound data demonstrate that BFLPD can reliably improve diagnostic accuracy while preserving data privacy, making it a promising tool with which smart cities can enhance their pandemic disease diagnosis capabilities.

Keywords

• blockchain
• federated learning
• pandemic disease
• smart cities
• COVID-19
• privacy-preserving