Cybersecurity threats in IoT-enabled smart agriculture systems: A SOC-centric detection and mitigation framework

Abstract

The emergence of cyber-physical systems in agricultural production has created unprecedented opportunities for precision farming while simultaneously introducing substantial security vulnerabilities. This research addresses the critical challenge of distinguishing sensor anomalies from malicious cyber-attacks in resource-constrained agricultural Internet of Things (IoT) environments. We propose a comprehensive Security Operations Center (SOC)-centric framework integrating signature-based detection, machine learning-based anomaly analysis, and novel fault-aware modeling within a multi-layered architecture specifically optimized for agricultural deployment scenarios. The framework's core innovation is a hybrid mathematical detection model that jointly analyzes anomaly scores and sensor fault probabilities, enabling differentiation between environmental sensor anomalies and cyber-attack signatures. Explainable Artificial Intelligence (XAI) modules utilizing SHAP (SHapley Additive exPlanations) and LIME (Local Interpretable Model-agnostic Explanations) enhance transparency in automated security decisions, addressing operational trust requirements in real-world farm deployments Experimental validation through containerized testbed simulations demonstrates exceptional performance: detection accuracy of 97.8%, precision of 96.4%, recall of 98.2%, F1-score of 0.972, and ROC-AUC of 0.989. The Fault-Attack Detectability Score (FADS) metric achieves mean performance of 0.910, indicating superior fault-attack differentiation capability. Comparative analysis reveals substantial improvements: 18.5% accuracy advantage over traditional intrusion detection systems, 12.3% improvement over machine learning autoencoder baselines, and 8.1% F1-score gain over random forest ensemble approaches. Resource efficiency analysis demonstrates practical viability for constrained agricultural deployments: 32% memory reduction, 87% network bandwidth reduction via edge processing, and 86% power consumption reduction. The framework provides a systematic, scalable, and explainable security solution that addresses the critical research gap between sensor health monitoring and cybersecurity management in Agriculture 4.0 systems.