Improving federated learning protection with digital envelopes

Abstract

The Federated Learning method was developed to to provide an alternative for the recent concerns with data privacy in machine learning. This method involves multiple parties to privately train local machine learning models with their own data, sharing with the global server only the models’ parameters that will be averaged to update the global model. Although private, such environments are constantly at the risk of suffering cyber-attacks that can compromise the information used in the process and/or the complete machine learning training. This work investigates the application of Digital Envelopes combined with Federated Learning, to improve protection against attacks to either the clients and the server.