SDN (Software-Defined Network) is a network architecture that separates the control and data planes, facilitating centralized control and dynamic management of network resources. Distributed Denial of Service (DDoS) attacks represent a critical challenge in SDN environments, as they flood the centralized controller with excessive malicious traffic. This overwhelms its computational resources and disrupts its ability to make accurate and timely network decisions. These attacks can result in network congestion, poor performance, and even network outages because SDN relies on a single controller to control traffic flow across the network. Metaheuristic algorithms, such as Particle Swarm Optimization (PSO), are optimization approaches that aim to find approximate solutions to complex problems that may not be susceptible to standard methods. This study evaluates PSO-based feature selection integrated with XGBoost classification for DDoS detection in SDN using the InSDN and Mininet datasets. The results demonstrate that XGBoost outperforms traditional classifiers, including Logistic Regression and Naïve Bayes, achieving over 99% accuracy, recall, and F1-score. The comparative analysis highlights XGBoost’s superior performance in minimizing false positives and false negatives, ensuring higher detection reliability. The AUC score of 0.99 further confirms its robustness in distinguishing attack and benign traffic.