Improved DDoS Attack Detection Through PCA-Enhanced Ensemble Machine Learning Approach

Distributed Denial-of-Service (DDoS) attacks pose a serious threat to network infrastructure by overwhelming systems with traffic from multiple sources, leading to service disruptions. Detecting such attacks in real time remains a complex challenge in cybersecurity, especially as reliance on digital platforms and financial transactions continues to grow. This study explores the effectiveness of combining Principal Component Analysis (PCA) with machine learning algorithms to detect and classify DDoS attacks using network log data. PCA, a dimensionality reduction technique, is employed to streamline data processing, enhance model efficiency, and mitigate overfitting. By reducing data complexity, PCA helps improve both the speed and accuracy of machine learning models. Several classifiers are evaluated in this research, including Logistic Regression, K-Nearest Neighbours (KNN), Support Vector Machines (SVM), Naïve Bayes, Decision Trees, Random Forest, AdaBoost, and XGBoost. Among these, ensemble methods, particularly when integrated with PCA, demonstrated superior performance. Experimental results revealed a detection accuracy of 97.86%, highlighting the potential of this hybrid approach. This high accuracy is particularly valuable in handling high-dimensional datasets commonly found in network traffic analysis. The findings underscore the critical role of dimensionality reduction in improving the effectiveness of cybersecurity solutions driven by machine learning, making this integration a promising direction for real-time DDoS detection and network protection.