Explore the critical aspects of data privacy in federated learning through this 35-minute conference talk by Xiaokui Xiao. Delve into the fundamentals of federated learning, its operational mechanisms, and the challenges it presents to data privacy. Examine various approaches to preserving privacy, including local gradient methods, multi-party computation (MPC), trusted hardware, and differential privacy. Analyze experimental results and real-world examples, such as age distribution of customers, to understand the practical implications of these techniques. Investigate the concept of model privacy in vertical federated learning and discuss potential mitigation strategies. Conclude with insights into future work, including the development of privacy frameworks and new techniques, while addressing other pertinent issues in the field. Gain valuable knowledge from this presentation delivered at the Association for Computing Machinery (ACM) conference, with the speaker representing the National University of Singapore. Read more