Machine Learning for Cybersecurity Cookbook

Federated learning

In this recipe, we will train a federated learning model using the TensorFlow federated framework.

To understand why federated learning is valuable, consider the next word prediction model on your mobile phone when you write an SMS message. For privacy reasons, you wouldn't want the data, that is, your text messages, to be sent to a central server to be used for training the next word predictor. But it's still nice to have an accurate next word prediction algorithm. What to do? This is where federated learning comes in, which is a machine learning technique developed to tackle such privacy concerns.

The core idea in federated learning is that a training dataset remains in the hands of its producers, preserving privacy and ...