2026-04-17T21:52:04.050Z http://ezaposleni.singidunum.ac.rs/rest/sciNaucniRezultati/oai

ezaposleni.singidunum.ac.rs/rest/sciNaucniRezultati/oai:3:9089 2024-12-03T20:10:47Z 3

XGBoost Hyperparameters Tuning by Fitness-Dependent Optimizer for Network Intrusion Detection, Chapter in LNNS Lecture Notes in Networks and Systems: ICCIS 2021: Communication and Intelligent Systems, Springer, volume 461 2022 http://ezaposleni.singidunum.ac.rs/rest/sciNaucniRezultati/oai/record/3/9089 https://link.springer.com/chapter/10.1007/978-981-19-2130-8_74 M. Zivkovic L. Jovanovic M. Ivanovic N. Bacanin I. Strumberger P. Joseph Network intrusion detection systems are frequently utilized for attack detection and network protection. However, one of the frequent issues intrusion detection systems face is the false positive detections. The research proposed in this paper suggests the hybrid FDO-XGBoost approach for tackling this issue, and according to the extensive simulations and comparative analysis against other common approaches, the proposed method is capable of achieving higher overall classification accuracy when compared to pure XGBoost, random forest and others. The FDO is utilized for adaptive search of the optimal architecture of the XGBoost. Proposed method is validated against widely used NSL-KDD benchmark dataset. The experimental findings indicate that the suggested FDO-XGBoost approach significantly outperforms other approaches in terms of accuracy, having precision and recall with average values of 0.82 and 0.77, respectively. Our method also achieved higher detection rate with all five tested attack classes (normal, probe, Dos, U2R, R2L), slightly outperforming PSO-XGBoost by 1–2% but with a much noticeable difference when it comes to other similar methods. bookPart Springer, Singapore 947 962 10.1007/978-981-19-2130-8_74 LNNS Lecture Notes in Networks and Systems: ICCIS 2021: Communication and Intelligent Systems, volume 461