The quick adoption of Internet of Things networks has highlighted critical security vulnerabilities, necessitating efficient intrusion detection mechanisms. Tackling the challenges inherent in resource-constrained IoT environments and highly imbalanced datasets, this paper proposes RAISE, a resilient anomaly-based intrusion detection framework. RAISE systematically incorporates SMOTE-based class balancing alongside gradient boosting classifiers to enhance detection accuracy, particularly for minority attack categories. Extensive evaluation using the tailored version of Bot-IoT dataset demonstrates that the proposed RAISE framework achieves improved precision, F1-scores, accuracy and recall across binary as well as multiclass classification tasks. The findings highlight RAISE’s potential to fortify IoT infrastructures against evolving cyber threats with minimal computational overhead.