BlockFedML: Blockchained Federated Machine Learning Systems
Shufen Wang · 2019
With the emphasis on security and privacy, Federated Machine Learning (FML) systems have become a research hotspot due to it can perform machine learning models without compromising security and privacy. However, there are two crucial challenges. One is a gradient information leak, and the other is vulnerable to integrity attacks. In this paper, we proposed a Blockchained Federated Machine Learning System, which called "BlockFedML." In BlockFedML, we develop Security Parameter Aggregation Mechanisms, Checkpoint based-Smart Contracts, Incentive Mechanisms, and Transfer Learning. Finally, we outlined the BlockFedML system and its applications and explained our future work.