Blockchain-Enabled Federated GRU-Based Secure Digital Twin Architecture for Smart Agriculture Recommendation Systems

Abstract

The continuous digitization of the modern farming sector demands secure, intelligent, privacy-preserving, and scalable infrastructures for real-time data analysis. However, existing smart farming systems face significant challenges, including cyberthreats, data authenticity issues, and the need for reliable decision support. This article proposes a secure Digital Twin (DT) architecture for smart agriculture recommendation systems, integrated with Blockchain and Federated Gated Recurrent Units (FGRU). At the perception layer, IoT sensors monitor soil, crop, and environmental data, which is gathered by a Request Control Authority (RCA) and transmitted to local models. To ensure privacy, a GRU-based Federated Learning (FL) approach is employed to detect cyberattacks—such as Sybil, Man-in-the-Middle (MITM), DDoS, and Replay attacks—without exposing raw decentralized data. Furthermore, a Blockchain-assisted Zero-Knowledge Proof-based Authority (ZKPA) mechanism is integrated to ensure data authenticity. The validated farming data is stored at the architecture’s final layer, enabling a Physical Twin to monitor real-time processes and generate precise recommendations. The architecture was evaluated using a paddy field dataset (26 features, 10,081 samples). Experimental results show that the proposed federated GRU model achieves perfect detection performance for all considered attacks, while the ZKPA-based authentication mechanism achieves a 98–99% authentication success rate with sub-10 ms verification time and only 15–25% additional computational overhead, which is better than existing works.