An Ensemble Feature Selection Method for IoT IDS
Alaa Alhowaide, Izzat Mahmoud Alsmadi, Jian Tang · 2020
The usage of the Internet of Things (IoT) is growing in both classical and new application domains. Such growth comes with a significant increase in cyber threats accompanied by delays in detection. Most of Intrusion Detection Systems (IDSs) do not employ feature selection approaches to reduce the data dimensionality. This research applied several feature selection methods and detection models on four different datasets. We also proposed an ensemble-based reliable feature selection method. Additionally, we proposed a novel mechanism to nominate features dynamically. Notably, the proposed ensemble-based feature selection methods reduced the evaluated dataset with 66% confidence. Furthermore, the proposed approach showed a noticeable improvement in feature selection methods' efficiency. Furthermore, the gained efficiency did not compromise the detection models' performance.