This study addressed the growing need for accurate and efficient intrusion detection by developing an IDS model that integrates the Dragonfly Algorithm (DA) for feature selection with XGBoost and Naïve Bayes classifiers. Using the UNSWNB15 dataset, the DA effectively reduced dimensionality while retaining essential features, enhancing the performance of the classification stage. Among the evaluated models, the DA-XGBoost configuration achieved outstanding results, including 100% detection accuracy, underscoring the strength of combining metaheuristic feature selection with advanced machine learning. The findings confirm that DA significantly contributes to reducing computational complexity without compromising accuracy. Moreover, the approach demonstrates robustness in handling real-world traffic patterns and attack types. These results position the proposed IDS framework as a viable candidate for deployment in real-time network security systems. Future work may extend this approach by applying it to other cybersecurity datasets or integrating it with deep learning methods to further enhance detection capabilities in evolving threat landscapes.
AL-ALLAWEE et al. (Thu,) studied this question.