SecuFarmArch: AI-Driven Cybersecurity Architecture for Smart Farming Using Digital Twins and Machine Learning

The integration of the Internet of Things (IoT), artificial intelligence (AI), and information and communication technology (ICT) in agriculture has significantly enhanced productivity, efficiency, and cost management. However, it has also introduced numerous cybersecurity vulnerabilities, posing substantial risks to the integrity of smart farming operations. Recognized by the Department of Homeland Security as critical to national security, protecting this infrastructure demands continuous threat modeling and vigilant monitoring. This paper proposes a robust cybersecurity framework leveraging AI, IoT, and Digital Twins (DTs) with a specific focus on mitigating tampering attacks. We employ machine learning models, particularly a multilayer perceptron(MLP)-based approach, trained on an IoT network traffic dataset to detect malicious patterns. Our results demonstrate the MLP’s superiority over traditional methods like K-Nearest Neighbors (KNN) in threat detection. Additionally, we highlight the innovative potential of DT technology for ongoing system auditing, enhancing immediate threat response, and forecasting potential vulnerabilities. This framework not only detects anomalies but also optimizes and secures agricultural systems, marking a significant advancement in agricultural technology. The contributions of this work are: i) a comprehensive framework for smart farming security focusing on real-time threat monitoring, ii) a Machine Learning (ML) Intrusion Detection System (IDS) for proactive threat detection, and iii) an innovative DT-based ongoing auditing system.