Intrusion Detection System for Smart Agriculture Using Deep Learning

Abstract: Smart agriculture has rapidly adopted the use of Internet of Things (IoT) technologies that enhance the process of monitoring farm conditions and making decisions based on that. Nevertheless, IoT devices deployed in open and harsh environments are very susceptible to DDoS attacks and other forms of cyber intrusions. This project aims at developing a solution for intrusion detection in smart agriculture systems by applying a deep learning approach. Specifically, this work focuses on the development of an IDS based on a hybrid architecture involving BiGRU and LSTM architectures in order to perform analysis of sequence data and identify malicious operations. To achieve this goal, the intrusion detection system will be built as a web application implemented in Python Flask. The application allows uploading of datasets, training of the model and visualizing the results. TBPTT will be applied to optimize model training process. In this paper, we assess the performance of our model based on metrics such as accuracy, precision, recall and F1-score. Moreover, it should be noted that we calculate mapping between attack severity level and agricultural impact indicators such as water use, fertilizer efficiency and crop risk. Our experiments show promising results regarding accuracy of the model.

Keywords: Smart Agriculture, Internet of Things (IoT), Intrusion Detection System (IDS), Deep Learning, BiGRU, LSTM, TBPTT, Cybersecurity, DDoS Attack Detection, Network Security, Flask Web Application, Time-Series Analysis, Agricultural Impact Analysis1