Comparison of Single and Ensemble Intrusion Detection Techniques using Multiple Datasets

The advancement of Internet of Things (IoT) technology raises numerous security concerns, as new threats emerge every day. Prior to preventing these threats, they must be detected. This makes intrusion detection a major priority. However, datasets play a significant role in intrusion detection. The dataset used to evaluate machine learning-based solutions has an effect on their accuracy. Most of the time, these datasets do not accurately reflect real network traffic and contains lots of redundant and irrelevant features that undermine Intrusion Detection System (IDS) efficiency. Motivated by the above, our work focuses on extracting the most relevant features from four datasets namely CICIDS2017, IoTID20, NSL-KDD and N-BaIoT datasets using information gain approach. Then we evaluated and compared some single and ensemble classifiers based four important performance metrics. Finally, these algorithms were combined in an ensemble learner to see how well they performed. Our findings are considered to be relevant in the combination of strong classification algorithms in the development of IDS systems and experimental results indicates that feature selection can yield better accuracy.