Multi-Layer Intrusion Detection in Smart IoT Systems Using Voting Classifier and Explainable AI with Flask Integration
Keywords:
Internet of Things, Industrial Internet of Things, security, cyber-attacks, intrusion detection, machine learning, lightweight, data balancing, cross-dataset transfer learning”.Abstract
Abstract—The fast rise of IoT and IIoT settings has raised the need for lightweight, scalable and intelligent intrusion detection systems. In this work, we propose an exhaustive ML and DL-based Intrusion Detection System and evaluate it with respect to ToN_IoT, WUSTL-IIoT 2021 and Edge-IIoTset datasets. The data preprocessing techniques used are outlier removal, Chi square feature selection, imbalance handling using SMOTE and undersampling. Several algorithms were implemented including CNN , DNN, DT , RF , LightGBM , Bagging , stack ensemble using RF, LightGBM and DT , and voting ensemble of Boosted DT, Bagging RF and XGBoost . The results of the experiments show the good performance of CNN (93.8%), DNN (94.4%), RF (96.3%), DT (96.6%), and the best performance of LightGBM (98.4%, 98.9%, and 96.7%) for the voting ensemble on ToN_IoT, WUSTL-IIoT 2021, and Edge-IIoTset respectively. Explainable AI techniques, like SHAP and LIME, offer feature level explainability. The deployment with Flask and SQLite enables safe, real time intrusion detection, visualization and monitoring over distributed industrial networks worldwide.