IoT-based Smart Agriculture System for Precision Farming

Abstract

The integration of Internet of Things (IoT) technologies in agriculture has led to the emergence of smart farming systems that enable real-time monitoring, data collection, and precision decision-making. However, the growing complexity of these interconnected systems presents two major challenges; the need for accurate agricultural prediction and the risk of cybersecurity threats targeting vulnerable IoT devices. To address these challenges, this work proposes two machine learning-based solutions a Random Forest-assisted wrapper (RFA-Wrapper) method for predicting crop productivity using time-series and environmental data, and AgriStackIDS, an ensemble intrusion detection system for securing smart agriculture networks. Both models were evaluated on real-world datasets using key performance metrics, including accuracy, F1-score, and recall. The results show that RFA-Wrapper outperforms traditional models in prediction tasks, while AgriStackIDS achieves high detection performance in both binary and multi-class classifications. These contributions demonstrate the effectiveness of machine learning in enhancing both productivity and cybersecurity in IoT-based agricultural environments.