Machine Learning and Privacy

This lecture covers the fundamental concepts of machine learning and privacy, focusing on the confidentiality of training data, privacy concerns during the machine learning life cycle, and attacks on private data used to train models. It delves into membership inference attacks, gradient inversion, differential privacy, and the trade-offs between utility and privacy in federated learning. The lecture also discusses the challenges of differential privacy in small datasets and the disparate impact in federated learning, emphasizing the importance of protecting sensitive training data to prevent privacy breaches.