A Scalable Deep Learning-Based Intrusion Detection System for Real-Time Cybersecurity in IoT Networks of the Sugar Industry

Abstract

The increasing integration of IoT-enabled systems in the sugar industry has enhanced operational efficiency but also exposed critical infrastructures to cyber threats. This paper presents the design and implementation of a scalable, real-time Intrusion Detection System (IDS) using deep learning-based Stacked Ensemble models, including CNN-LSTM, AE-RF, XGB-DNN and LSTM-Bagging, to detect and mitigate severe cyber-attacks such as DDoS, Ransomware and Man-in-the-Middle (MITM). The IDS monitors operational parameters like temperature, pressure, energy consumption and flow across key divisions of the sugar manufacturing process, including Cane Milling, Cane Preparation, Cane Handling, Power Conversion and Bagasse Conveying systems. The proposed system classifies attack patterns and anomalies with high accuracy, precision, recall and F1 score. It is designed to operate efficiently in large-scale, resource-constrained IoT environments. Furthermore, automated threat mitigation strategies ensure minimal disruption to industrial processes. The results demonstrate the IDS's effectiveness in safeguarding IoT networks in the sugar industry by promptly identifying and addressing critical cyber threats.