PeerJ Computer Science, Год журнала: 2025, Номер 11, С. e2745 - e2745

Опубликована: Март 17, 2025

As the number of connected devices and Internet Things (IoT) grows, it is becoming more important to develop efficient security mechanisms manage risks vulnerabilities in IoT networks. Intrusion detection systems (IDSs) have been developed implemented networks discern between regular network traffic potential malicious attacks. This article proposes a new IDS based on hybrid method metaheuristic deep learning techniques, namely, flower pollination algorithm (FPA) neural (DNN), with an ensemble paradigm. To handle problem imbalance class distribution intrusion datasets, roughly-balanced (RB) Bagging strategy utilized, where DNN models trained by FPA cost-sensitive fitness function are used as base learners. The RB derives multiple training subsets from original dataset proper weights incorporated into attain unbiased models. performance our evaluated using four commonly utilized public NSL-KDD, UNSW NB-15, CIC-IDS-2017, BoT-IoT, terms different metrics, i.e., accuracy, precision, recall, F1-score. results demonstrate that outperforms existing ones accurately detecting intrusions effective handling problem.

Язык: Английский