Development of Hybrid Intrusion Detection System Leveraging Ensemble Stacked Feature Selectors and Learning Classifiers to Mitigate the DoS Attacks

Denial of service (DoS) attacks occur more frequently with the progressive development of the Internet of things (IoT) and other Internet-based communication technologies. Since these technologies are deeply rooted in the individual’s comfort life, protecting the user’s privacy and security against the growing DoS attack has become a major challenge among researchers. In recent times, intrusion detection systems (IDS) have developed a vital part in ensuring security against these growing attacks. IDS is still unable to attain the optimum categorization performance due to a few bottlenecks. The speed and performance of the existing IDS are challenged by the intricacy of high-dimensional data and the efficacy of the conventional base classifiers. To tackle this aforementioned problem, this research article presents the hybrid IDS based on the combination of stacked feature selection methods such as Random Boruta Selector (RFS), Relief, Pearson coefficient (PCE) and Stacked learning classifiers (SLF). To reduce the dimension of the data features and to select the optimal feature sets, novel integration of RFS, Relief, PCE are deployed. As the final step, stacked classifiers are used for the classification of DoS attacks. All the trials in this framework were accompanied utilizing CICDDoS-2019 datasets and contrasted with the other similar models. The validation boundaries such as accuracy, precision, recall, specificity, and F1-score are used to evaluate the proposed framework. With an F1-score of 96%, accuracy of 96.5%, precision of 96.0%, and recall of 95.8%, the suggested model obtained a CICDDoS-2019 score of 96%. Compared with the other traditional classifiers, the suggested framework has produced the best classification performance in detecting the DoS attacks.