2026-05-05T12:14:04.955Z http://ezaposleni.singidunum.ac.rs/rest/sciNaucniRezultati/oai

ezaposleni.singidunum.ac.rs/rest/sciNaucniRezultati/oai:1:11780 2026-01-06T17:11:37Z 1

Hyperparameter Tuning with Modified Metaheuristics for Intrusion Detection in IoT Networks 2025 http://ezaposleni.singidunum.ac.rs/rest/sciNaucniRezultati/oai/record/1/11780 https://ieeexplore.ieee.org/document/11314269 I. Kosta S. Anetic L. Jovanovic M. Antonijevic M. Zivkovic N. Bacanin Intrusion detection in Internet of things (IoT) networks is critical for safeguarding devices and communication infrastructures. This study investigates IoT intrusion detection using XGBoost classifiers optimized via multiple metaheuristic algorithms and introduced a modified version of a recently proposed optimization metaheuristic. Hyperparameter tuning was performed across thirty independent runs, and models were evaluated using standard evaluation metrics. Results demonstrate that the proposed modified optimizer helps tune extreme gradient boosting (XGBoost) classifier performance consistently so that it outperforms other evaluated approaches, achieving the highest objective function values, lowest error rates, and balanced detection of normal and anomalous traffic. Convergence analysis and error distributions further confirm the stability and reliability of this approach. These findings highlight that careful metaheuristic optimization of XGBoost hyperparameters can significantly improve intrusion detection performance in heterogeneous IoT environments. conferenceObject 1 4 10.1109/TELFOR67910.2025.11314269 2025 33rd Telecommunications Forum (TELFOR), IEEE, Belgrade, Serbia