Evaluating the Effectiveness of Various Model Combinations for Network Intrusion Detection on UNSW-NB15

Internet technology's advancement has rendered network security a paramount issue. In response, numerous computer network intrusion detection systems have been developed and deployed over recent years, utilizing diverse methodologies. The UNSW-NB15 dataset, a comprehensive realworld benchmark for network intrusion detection, is recognized for its varied attack types and comprehensive feature sets. This study encompassed a comparative analysis of ensemble, fusion, and stacking models employing the UNSW-NB15 dataset. Notably, the stacking model exhibited superior performance, boasting a test accuracy of <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{9 4. 4 3 \%}$</tex> in a multi-class classification setting. Our results underscore the stacking method's potential for enhancing detection efficacy and reducing false positives, with specific configurations scoring exceptional results across various assessment criteria.