How to Use Federated Learning to Protect Your Data

Federated learning by keeping all the data on the notes solves the most glaring security hole in your data, which is that you have to give the data in plain text to someone else. Vanilla federated learning stops there, though. It does not address the problem that we talked about also at the start,. In principle, you can invert a machine learning model and figure out things that were true of the training data. So federated learning isn't going to help with that. You are in general protected from other nodes in the network. That's because all communication, all your model is only ever shown to the server. But if privacy is concern, then there are additional steps you might