<p>Detecting intrusions in wireless sensor networks (WSNs) is crucial for maintaining data security. However, the inherent class imbalance in intrusion datasets often degrades the performance of traditional machine learning algorithms. This study presents a novel hybrid metaheuristic approach, integrating the whale optimization algorithm (WOA) and differential evolution (DE), termed HWOADE, for optimal feature selection in imbalanced intrusion detection systems. To address class imbalance, a modified ADAptive SYNthetic (MADASYN) method is applied for oversampling. The proposed HWOADE algorithm enhances feature selection and improves classification accuracy by leveraging a support vector machine classifier with a radial basis function kernel. Evaluated on NSL-KDD and UNSW-NB15 datasets, the proposed model achieves notable results, with an accuracy of up to 97.6% on the NSL-KDD dataset and a precision and recall of 95.4 and 90.4%, respectively. Compared to baseline models, the HWOADE method consistently outperforms in terms of F1-score, accuracy, and precision. This approach significantly contributes to improving the robustness of machine learning-based intrusion detection systems, ensuring better data security in WSNs.</p>