Federated Edge-to-Cloud Analytics: A Privacy-Preserving Real-Time Big Data Processing Framework for IoT-Driven Smart Ecosystems
Abstract
Traditional cloud-centric IoT analytics suffers from unacceptable delays, large bandwidth consumption, and potential privacy leakage; standard Federated Learning (FL) approaches addresses privacy leakage but introduce communication overhead and ignore the timing constraints imposed by heterogeneous IoT edge networks. We propose Federated Edge-to-Cloud Analytics (FECA), an adaptive hierarchical edge-cloud orchestration framework for low-latency, privacy-preserving analytics in distributed smart ecosystems. FECA employs Dual-Axis Urgency-Sensitivity Routing (DA-USR) to dynamically evaluate telemetry at the edge, directing latency-critical inference to local gateways while routing sensitive updates through a multi-tier federated hierarchy. A calibrated (ε,δ)-Differential Privacy layer is embedded in local training to defend against gradient inversion attacks with minimal performance loss. Evaluated on smart city, healthcare, and industrial automation benchmarks, FECA achieves 38.4 ms mean edge-inference latency (68.4% reduction vs. centralized cloud processing) and 74.6% lower uplink overhead than flat FedAvg. Under a strict privacy budget (ε=2.0), FECA attains 89.4% F1-score, limiting accuracy degradation to 3.2% while accelerating convergence by 58.8%.