Securing IoT-Enabled Smart Agriculture through Blockchain-Enhanced Federated Anomaly Detection: An Empirical Evaluation

IoT-enabled smart agriculture is reshaping farm management through continuous sensing, remote monitoring, irrigation automation and data-supported decision-making. These benefits also expose farm cyber-physical systems to attacks that can falsify device identity, replay legitimate messages, inject abnormal packets, alter data streams, interrupt services and permit unauthorised access. This study develops and evaluates B-FedAgriIDS, a blockchain-enhanced federated anomaly-detection framework for securing smart-agriculture IoT environments. The framework places autoencoder-based anomaly detection at edge gateways, applies federated averaging to support collaborative learning without centralising raw farm data, and uses a permissioned blockchain to manage device identity, access verification, alert integrity and auditable logging. A comparative empirical design was used to assess a centralised autoencoder baseline, a federated-only autoencoder and the proposed blockchain-enhanced federated model. The proposed model achieved the strongest performance, with 97% detection accuracy and a 4% false alarm rate, compared with 89% accuracy and 12% false alarms for the centralised baseline and 94% accuracy and 8% false alarms for the federated-only model. Blockchain operations remained within an estimated 50-100 ms latency range, which is suitable for monitoring, authentication and alert-recording functions in farm networks. The findings show that combining privacy-preserving learning with permissioned ledger governance can strengthen detection, accountability and operational trust in distributed digital-agriculture systems.