When Federated Learning Meets Blockchain: A New Distributed Learning Paradigm
share
Motivated by the advancing computational capabilities of wireless end user equipments (UEs), as well as the increasing concerns about directly sharing private data, a new machine learning paradigm, namely federated learning (FL) has emerged. By training models locally at UEs and aggregating trained models at a central server, FL is capable of avoiding data leakage from UEs, thereby helping to preserve privacy and security. However, the traditional FL framework heavily relies on the single central server and may fall apart if such a server behaves maliciously. To address this single point of the potentially failure issue, this work proposes a blockchain aided FL framework. The proposed framework can well prevent the malicious UEs from poisoning the learning process, and further provide a self-motivated and reliable learning environment for UEs. In addition, we investigate the critical issues in the proposed framework and provide experimental results to shed light on possible solutions.
READ FULL TEXT