Butterfly optimization-based ensemble learning strategy for advanced intrusion detection in internet of things networks

Abstract

The massive growth in internet of things (IoT) devices has led to enhanced functionalities through their interconnections with other devices, smart infrastructures, and networks. However, increased connectivity also increases the risk of cyberattacks. To protect IoT systems from these threats, intrusion detection systems (IDS) employing machine learning (ML) techniques have been developed to identify cybersecurity threats. This paper introduces a novel ensemble IDS framework called butterfly optimization-based ensemble learning (BOEL). This framework integrates the butterfly optimization algorithm (BOA) with ensemble learning techniques to improve IDS detection performance in IoT networks. BOEL is designed to accurately detect various types of attacks in IoT networks by dynamically optimizing the weights of base learners, which are the four sophisticated ML gradient-boosting algorithms (GBM, CatBoost, XGBoost, and LightGBM) for each attack category, and identifying the best weight combination for ensemble models. Experiments conducted on two public IoT security datasets, CICIDS2017 and Bot-IoT, demonstrate the robustness of the proposed BOEL in intrusion detection across diverse IoT environments, achieving 99.795% accuracy on CICIDS2017 and 99.966% accuracy on Bot-IoT. These results outline the successful application of diverse learning approaches and highlight the framework’s potential to enhance IDS in addressing IoT cyber threats.