A Federated Meta Learning-Based Secure Data Consolidation Scheme for Industrial AIoT Leveraging Drone

Abstract

Amidst the technological revolution, the convergence of Industrial Artificial Intelligence of Things (Industrial AIoT) signifies a profound transformation in industrial operations. Nonetheless, persistent concerns revolve around data privacy, security, and connectivity challenges. Drones emerge as pivotal aids for Industrial AIoTs, particularly in areas with limited connectivity. While Federated Learning (FL) and Meta-Learning (ML) address data privacy and adaptability, challenges like data heterogeneity, scarcity, model positioning, unauthorized data tampering, and cyber threats endure. To tackle these issues, this paper presents a Federated Meta-Learning (FML)-based secure data consolidation scheme, utilizing drones for data consolidation, especially in remote, poorly connected regions, followed by secure blockchain storage. It incorporates an Information Gain Ratio (IGR)-based feature selection method to manage data diversity, a two-phase authentication system merging XOR filtering and Chronological Nonce Authentication for entity validation, and secure model consolidation using Hampel filters and performance checks to validate model updates. A real-world proof of concept demonstrates superior performance compared to state-of-the-art literature.