Research Square (Research Square), Journal Year: 2024, Volume and Issue: unknown

Published: Dec. 30, 2024

Network traffic anomaly detection involves the rapid identification of intrusions within a network through detection, analysis, and classification data.The variety cyber attacks encompasses diverse attack principles. Employing an indiscriminate feature selection strategy may lead to neglect key features highly correlated with specific types. This oversight could diminish recognition rate for that category, thereby impacting overall performance model.To address this issue, paper proposes model based on fusion attack-dimensional features. Firstly, construct binary datasets independently each class perform individual extract positively class. The are then fused by employing combination methods. Subsequently, sub-datasets, base classifiers trained. Finally, ensemble learning approach is introduced integrate predictions classifiers, enhancing robustness model.The proposed approach, validated NSL-KDD UNSW-NB15 benchmark datasets, outperforms latest methods in field achieving \(2%\) \(7%\) increase precision weighted averages.

Language: Английский