Federated Sampling for Privacy-Preserving and Efficient Data Sharing in IoT Networks

This paper presents a federated sampling strategy aimed at enhancing data efficiency and privacy in IoT networks. By introducing a novel approach to summarization and synthetic data generation, we enable IoT devices to share compact statistical representations of data rather than raw data, significantly reducing the amount of data transmitted while preserving privacy. Our evaluation, using a protocol classification task, demonstrates that synthetic data generated from these summaries effectively replicates the statistical properties of the original data. This proof-of-concept work lays the foundation for future advancements, including network-wide simulations, intrusion detection mechanisms, and federated learning-inspired aggregation techniques, paving the way for resilient and privacy-preserving IoT systems.