<p>The necessity of understanding machine learning in cybersecurity has increased dramatically. Extreme Gradient Boosting, a complex boosting algorithm, excels at cyber threat investigation and detection. However, hyperparameter optimization determines the success of Extreme Gradient Boosting. Hyperparameter optimization using traditional methods is time-consuming and imperfect. The nature-inspired Dragonfly Algorithm is used to optimize the hyperparameters of Extreme Gradient Boosting in this work. The Dragonfly Algorithm balances search space exploration with exploitation, similar to dragonflies’ static and dynamic swarming. This approach improves the detection and evaluation of cyber attacks. The proposed method achieved 97.1% accuracy, 0.92 sensitivity, 0.981 precision, 0.991 F1-score, and 0.942 recall on the Industrial Internet of Things dataset. The results demonstrate the performance of the method with and without hyperparameter tuning and compare it with various benchmark machine learning models. The proposed method outperforms standard optimization methods in detection accuracy, precision, recall, and F1-score on benchmark cybersecurity datasets. The authors demonstrate that the Dragonfly Algorithm is efficient in processing resources and convergence speed, making it a promising choice for real-time cyber threat detection systems.</p>