As Internet of Things (IoT) is increasingly using in various fields, attackers are taking advantage of this potential for IoT devices to expose the users’ data integrity and security. Especially, they are seriously threatened by a distributed denial of service (DDoS) attacks. A DDoS attack aims to shut down the central server by overloading it with numerous requests until it reaches its capacity. Deep learning-based approaches can spot such dynamic attacks. This research proposes a machine learning-based lightweight boosting classification approach for an effective IoT-DDoS attack detection model based on the resampling techniques to minimize the data imbalance in the IoT security dataset with hybrid feature selection using principal component analysis (PCA) and Recursive Feature Elimination methods. The proposed approach performance is compared with six tree-based models, namely, Decision Tree, Random Forest, AdaBoost, CatBoost, XGBoost, and LightGBM in terms of time, ROC ACU, accuracy, MAE, RMSE and F1-score. The experimental results showed that the model with the lowest error rate on the resampled dataset with highest accuracy scores (99.97%) is Random Forest. However, the model with the shortest runtime on the resampled dataset is LightGBM, with 3.54 s. LightGBM demonstrates superior time efficiency compared to many other models, making it an attractive choice for scenarios where rapid processing is critical.