Towards secure and reliable aggregation for Federated Learning protocols in healthcare applications

Federated Learning (FL) is an AI framework that enables collaborative and distributed training across multiple users to learn a global model while preserving the privacy of the data held locally at different sites. However, the aggregation process of FL that relies on a centralized server to update the global model parameters exposes the protocol to several vulnerabilities. Thus, the privacy and security concerns in FL systems need to be further investigated to fully leverage the capabilities of this protocol, especially in industries involving highly sensitive data, such as healthcare. As part of this study, We emphasize the security challenges in the FL systems and propose a conceptual solution for a secure and efficient FL protocol based on defensive and compression mechanisms, respectively. Our work hopes to properly highlight the susceptible adversaries and attacks that need to be considered. Furthermore, our proposal constitutes a significant step towards a reliable aggregation method specifically designed for healthcare.