Machine learning can help us diagnose and fight cancer, decide which school is the best for our children and make the smartest real estate investment. But you can only answer these questions with access to private and personal data, which requires a novel approach to machine learning. This approach is called Secure and Private AI and, in recent years, has seen great strides, as you will see in the following recipes.
This chapter contains the following recipes:
- Federated learning
- Encrypted computation
- Private deep learning prediction
- Testing the adversarial robustness of neural networks
- Differential privacy using TensorFlow Privacy
