ROSE-BOX: A Lightweight and Efficient Intrusion Detection Framework for Resource-Constrained IIoT Environments

The rapid advancement of the Industrial Internet of Things (IIoT) has transformed industrial automation, enabling real-time monitoring and intelligent decision making. However, increased connectivity exposes IIoT systems to sophisticated cyber threats, which may pose significant security risks, especially in resource-constrained IIoT environments where computational efficiency is critical. Existing intrusion detection solutions often suffer from high computational overhead and inadequate adaptability, rendering them impractical for real-time deployment in IIoT environments. To address these challenges, this study introduces a lightweight and efficient intrusion detection framework tailored for resource-constrained IIoT environments. Firstly, an XGBoost-assisted Random Forest (XGB-RF) method is proposed to select the most important features to obtain an optimal feature subset. Moreover, SMOTE (Synthetic Minority Oversampling Technique) is utilized to balance the optimal feature subset to improve detection precision. Furthermore, to reduce computing resource requirements and latency while improving detection performance, Bayesian optimization is applied to fine-tune the parameters of XGBoost (BO-XGBoost) to obtain the best detection results. Finally, extensive experiments on benchmark datasets, including CIC-IDS2017, CSE-CIC-IDS2018, and CIC-DDoS2019, demonstrate that the proposed method, which we call ROSE-BOX (Random Forest, Synthetic Minority Oversampling Technique, and BO-Xgboost), achieves a detection accuracy exceeding 99.85% while maintaining low latency and CPU occupancy rates. Our findings highlight the robustness, lightweight nature, and efficiency of ROSE-BOX, making it well-suited for real-time intrusion detection in resource-constrained IIoT environments.