AgroCyberShield: cybersecurity strategies for protecting IoT-driven smart farming networks

As more IoT-powered Smart farming systems push automation to the next level, many farms are facing security risks, including spoofing, data breaches, and denial-of-service (DoS) attacks. Most current intrusion detection methods do not perform well due to limited scalability and the decision system's limited interpretability, as well as the dynamic nature of agricultural settings. On the one hand, traditional machine learning models fail to capture more complex temporal-spatial patterns in network traffic; on the other hand, many existing deep learning-based systems lack satisfactory explainability and are not feasible for edge deployment. To overcome these limitations, this paper offers AgroCyberShield, a new cybersecurity framework for IoT-enabled smart agriculture. At the core of this framework is the proposed FarmSecureNet algorithm a hybrid deep learning model that integrates Convolutional Neural Networks (CNNs), Bidirectional Long Short-Term Memory (BiLSTM), and attention for effective and explainable intrusion detection. These two datasets come from different sources on which the model is trained to enable generalisation. Extensive experiments demonstrate that FarmSecureNet achieves 96.4% classification accuracy, with F1-score, precision, and recall outperforming baseline models. We present further validation of the framework with SHAP-based explainability, confusion matrix analysis, and ablation studies. Moreover, the low latency and minimal memory footprint, as demonstrated in edge-deployment tests on Raspberry Pi and Jetson Nano, further support real-time performance. The developed AgroCyberShield framework enhances the cybersecurity of smart farms while improving decision transparency, contributing to the sustainable and safe digitisation of agriculture.