Federated Learning - Schemes, Privacy, and Security

Explore the challenges and solutions in federated learning through this 45-minute conference talk from BSides Edmonton 2025. Discover how federated learning has emerged as a privacy-preserving alternative to centralized machine learning, addressing critical issues such as data protection regulations, privacy concerns, and technical difficulties in data aggregation. Learn about the fundamental concepts of federated learning, where data owners collaboratively train global models under central parameter server coordination without transferring raw data. Examine the applications of federated learning across diverse fields including healthcare, agriculture, cybersecurity, and Internet of Things (IoT) implementations. Understand the key challenges facing federated learning systems, particularly poor convergence under heterogeneous conditions that can degrade global model performance. Investigate security vulnerabilities and data leakage risks that persist despite the privacy-enhancing design of federated learning frameworks. Gain insights into potential solutions for overcoming these technical and security limitations in distributed machine learning environments.