Federated Edge Intelligence: Enabling Privacy-Preserving AI for Smart Cities and IoT Systems

Abstract

Federated Edge Intelligence is an innovative data processing paradigm that empowers devices with artificial intelligence capabilities while preserving data privacy throughout the AI lifecycle. By exploring Federated Learning and Edge Intelligence for processing and communicating the smallest amount of data, we introduce a foundation architecture for Federated Edge Intelligence that is interoperable with existing spatial and temporal IoT data models. Architecting Federated Edge Intelligence requires careful attention to resource budget parameters, such as security, communication, and processing. To demonstrate the realizability and utility of our core idea, we present the usage of Federated Edge Intelligence for different Smart City and IoT case studies, spanning phenomena monitoring, self-aware actions, and system automation. Throughout the case studies, we increase the technology readiness level of Federated Edge Intelligence, demonstrating possibilities of both ambitious and low-tech achievements, utilizing powerful but also lightweight models and federated learning techniques.