An intelligent blockchain-based access control framework with federated learning for genome-wide association studies

Abstract

Genome-Wide Association Studies (GWAS) aim to find various variations in human disorders and have become one of the most commonly-used methods to find the pathogenesis and genetic mechanisms of complex diseases. However, the GWAS process needs to frequently search the genome-wide data, especially in the calculation process of multi-party participation. The statistical value calculation and interactive search of Single Nucleotide Polymorphisms (SNPs) and model training processes might easily disclose personal information. Therefore, to solve these problems, we propose a Blockchain-based access control Framework for GWAS with Federated Learning-BFGF. Specifically, before training local models, this framework implements Automated Quality Control (AQC) to guarantee the quality of training data. Design the authentication mechanism in blockchain to filter out users who are malicious attackers to protect the security of other users' information initially. To accelerate the speed of cloud model training and resist multiple attacks in federated learning, propose a periodic aggregation method combining differential privacy mechanisms. Finally, simulated experiments have shown that the BFGF framework can protect the security of genetic data and balance availability and accuracy.