Blockchain and FL-Based Secure Architecture for Enhanced External Intrusion Detection in Smart Farming

Smart farming influences advanced technologies to optimize agricultural procedures, yet it meets significant cybersecurity challenges, particularly in external intrusion detection (EID). This article proposes a novel architecture combining blockchain technology and federated learning (FL) to reinforce the security of smart farming systems (SMSs) against external threats. The integration of blockchain ensures data authentication and transparent data storage, while FL enables collaborative model training without compromising data privacy. Our architecture employs ensemble learning (EL) for the local model at the ensemble layer to train each smart land's data and offers privacy-prevented security. These devices utilize FL techniques to collaboratively train intrusion detection models while preserving the confidentiality of sensitive data. The aggregated model completes data aggregation at the authentication layer, and the Proof of Authentication Consensus Algorithm is leveraged for smart land's data authentication. The Internet of Things Sensor device's identical information of smart lands is stored at the macro base stations (MBSs). After downloading the aggregated values of the aggregated model, the local model transfers the smart lands information to the Cloud layer for decision making and decentralized storage. The validation outcomes of the proposed architecture demonstrate excellent performance, with an average processing time of 3.663 s and 0.9956 accuracy for smart land compared to existing frameworks.