Paper Key : IRJ************638
Author: Meenuga Pranaya Praharshitha,Dr. D William Albert
Date Published: 06 Apr 2025
Abstract
Machine Learning (ML) has proven to be a powerful approach for building Intrusion Detection Systems (IDS), particularly in the context of Internet of Things (IoT) environments where security is critical. This study explores the application of ensemble ML techniques to develop a time-efficient and highly accurate IDS specifically designed for smart IoT systems.The proposed system processes data collected from both network traffic and live IoT sensor inputs, with the primary goal of detecting and classifying various network-based attacks. Several machine learning modelsincluding Logistic Regression, Random Forest, Extreme Gradient Boosting (XGBoost), and Light Gradient Boosting Machine (LGBM)were evaluated using the DS2OS dataset. This dataset poses a significant challenge due to its highly imbalanced distribution of normal versus anomalous traffic.In this research, a novel intrusion detection model named LGB-IDS is introduced, built upon the LGBM algorithm. The effectiveness of this model was assessed using performance indicators such as accuracy, computational efficiency, error rate, true positive rate (TPR), and false negative rate (FNR). While both XGBoost and LGBM reached high accuracy levels of approximately 99.92%, LGBM showed superior speed and resource efficiency, making it more suitable for IoT devices with limited processing capabilities.Overall, the LGB-IDS model demonstrated a high detection rateexceeding 90%and maintained low false alarm rates with reduced processing time. These results emphasize the models potential for real-time intrusion detection in IoT networks, where both accuracy and efficiency are essential for maintaining cybersecurity.
DOI Requested