A survey of federated learning for Internet of things: recent advances, research problems and solutions

Federated learning, an emerging paradigm in machine learning, addresses challenges associated with privacy, communication costs, and scalability. Instead, only model updates are transmitted, preserving the privacy of sensitive information generated by IoT devices. This abstract explores the benefits of federated learning for IoT applications, emphasizing its potential to handle distributed and dynamic environments. The scalability of federated learning is highlighted as a key advantage, making it well-suited for the diverse and extensive networks characteristic of IoT deployments. Furthermore, the energy efficiency of federated learning contributes to its appeal for resource-constrained IoT devices, enabling on-device model training without compromising power resources. While federated learning introduces promising opportunities, challenges such as heterogeneity in devices, model convergence, and security considerations must be addressed. Ongoing research and development efforts are crucial for refining federated learning frameworks, ensuring their robustness, and facilitating widespread adoption in the rapidly evolving landscape of IoT applications.