Intrusion detection in IoT using ensemble approach

Abstract

Protection of the Internet of Things (IoT) devices is an area of concern, even with the success that has been achieved in this area. IoT involves configuring and deploying smart devices to send and share information. Some IoT devices carry sensitive information, which attracts the attention of cybercriminals. Intrusion detection systems have been widely proposed as one measure of defending networks against any malicious activities. This work proposes a stacked ensemble intrusion detection technique based on extreme learning machine (ELM), support vector machine (SVM), and KNeighbors (KNN) classifiers as base learners, and logistic regression (LR) as the meta-learning algorithm. Firstly, the dataset is cleaned and then grouped using the cross-validation procedure. Secondly, hyperparameter tuning of the algorithms is done using the grid search technique. Finally, with the tuned parameters, the classification job is done. The evaluation of the model is performed using the IoT_ToN network dataset. The performance of the proposed stacked ensemble method is compared with that of single algorithms. The obtained results clearly show the outstanding performance of the proposed stacked ensemble approach with respect to accuracy, precision, recall, and f1-score. The proposed model scored 96% across all the measured metrics outperforming the standalone algorithms. This study concludes that the stacked ensemble approach can potentially improve the performance of intrusion detection systems.