FEDERATED LEARNING FOR PRIVACY-PRESERVING ENGINEERING ANALYTICS

Abstract

Engineering analytics increasingly depend on data-driven insights derived from distributed data sources such as IoT devices, sensor networks, and industrial systems. Conventional centralized analytics approaches require the aggregation of raw data into a central server, raising privacy concerns and violating regulatory constraints. Federated learning (FL) has emerged as a paradigm that enables collaborative model training without sharing raw data, thereby facilitating privacy-preserving engineering analytics across distributed environments. This paper discusses FL architectures, privacy mechanisms, and engineering use cases. We propose a federated analytics framework tailored for engineering applications, evaluate its performance, and analyze privacy guarantees. Results from IoT and industrial datasets illustrate that FL can reduce privacy risks while maintaining analytical utility and scalability. Challenges and future research directions in privacy-preserving FL are discussed.