Federated Edge Intelligence for Secure and Scalable IoT Enabled Smart Systems

Abstract

This paper proposes a federated edge intelligence framework to effectively tackle some of the important issues of privacy, security and scalability in Internet of Things enabled smart systems. An effective approach that combines federated learning with edge intelligence while adding a trust aware aggregation mechanism to select benign updates so as to avoid contamination of the global model by malicious forces and improve its robustness. Secure aggregation and differential privacy preserve privacy, while edge based orchestration and model compression techniques improve scalability. The evaluation is done on heterogeneous IoT datasets, experimental network environments and performance metrics are calculated to measure accuracy, latency, communication overhead along with the security effectiveness. Results show that it performs with higher accuracy (97.3%), lower communication cost and much less energy cost and around 96% attack detection rate against baseline methods. It represents a good compromise between learning efficiency and system wide constraints. In summary, this work gives an efficient and versatile solution to enable secure, scalable and privacy preserving intelligence in distributed IoT environments.