Improving Internet of Things Cyber Attack Detection with Information Gain and Decision Tree

Abstract

The rapid proliferation of Internet of Things (IoT) devices within digital ecosystems has enhanced efficiency and availability but has also expanded the attack surface for cyber threats. This study aims to improve intrusion detection accuracy in IoT environments by addressing two key challenges: class imbalance and high feature dimensionality. Random Undersampling (RUS) is employed to mitigate data imbalance in the CIC IoT 2023 dataset, while feature selection is performed using the filter-based Information Gain method. A decision tree classifier is implemented and validated using k-fold cross-validation to ensure result reliability. Experimental results demonstrate that the proposed approach achieves an accuracy of 88.7%, outperforming a wrapper-based method, which attained 87.3%. These findings confirm that an appropriately designed filter-based feature selection strategy can effectively enhance the performance of intrusion detection systems for IoT security.