Federated Ensemble Learning for Privacy-Preserving Intrusion Detection in Industry 5.0-Enabled Smart Factories

Abstract

Industry 5.0-enabled smart factories integrate industrial Internet of Things (IIoT) devices, cyber-physical production systems, intelligent automation, collaborative robotics, and data-driven decision mechanisms. This interconnected environment improves manufacturing flexibility and operational intelligence but simultaneously enlarges the attack surface and creates substantial challenges for centralized intrusion detection. Conventional centralized approaches require sensitive operational data to be transferred to a common repository, creating privacy, bandwidth, governance, and single-point-of-failure concerns. This paper proposes a conceptual Federated Ensemble Learning (FEL) framework for privacy-preserving intrusion detection in Industry 5.0 smart factories. The framework combines federated learning with heterogeneous ensemble detection so that manufacturing sites can collaboratively improve detection models without directly sharing raw industrial traffic. Blockchain-inspired trust mechanisms are incorporated to support decentralized model accountability, while lightweight identity management is considered for resource-constrained IIoT environments. The framework is theoretically positioned around distributed learning, collaborative intelligence, privacy preservation, and resilient cyber-physical manufacturing. The analysis indicates that federated ensemble learning can provide a more appropriate architectural balance between local data sovereignty and cross-factory threat intelligence than purely centralized detection. However, communication overhead, non-identically distributed industrial traffic, poisoned model updates, computational constraints, and coordination complexity remain significant limitations. The proposed framework therefore represents a research-oriented architecture rather than a claim of experimentally validated performance. Its principal contribution is the integration of privacy-preserving collaborative learning and ensemble-based intrusion detection into an Industry 5.0 manufacturing context.