Akademik Veri Yönetim Sistemi
International Conference on Mathematics and Mathematics Education (ICMME-2025), İstanbul, Türkiye, 11 - 13 Eylül 2025, cilt.1, ss.103-104, (Özet Bildiri)
The increasing frequency and sophistication of Distributed Denial-of-Service (DDoS) attacks pose critical threats to network availability and security. This study aims to develop an effective classification model for DDoS detection by leveraging supervised machine learning algorithms. The CIC-DDoS2019 dataset was employed, consisting of over 400,000 traffic instances and 78 features, which were reduced to 32 through preprocessing operations, including correlation analysis and feature elimination. Six widely used algorithms—Random Forest, Support Vector Machine (SVM), k-Nearest Neighbors (KNN), XGBoost, LightGBM, and CatBoost—were implemented to classify normal and attack traffic.Model performance was evaluated using Accuracy, Recall, Precision, and F1 Score, supported by confusion matrix analyses. Experimental results revealed that CatBoost achieved the highest performance, with an Accuracy of 98.99%, Recall of 0.8233, Precision of 0.8505, and F1 Score of 0.8341. XGBoost (Accuracy: 98.85%, F1: 0.8184) and KNN (Accuracy: 98.85%, F1: 0.7957) also exhibited strong results, while SVM showed comparatively lower performance (Accuracy: 97.85%, F1: 0.6853). These findings highlight the importance of algorithm selection, as high accuracy values may mask low recall rates that can undermine real-world security effectiveness. Overall, the results demonstrate that supervised machine learning algorithms can effectively detect DDoS attacks, with CatBoost particularly excelling in handling class imbalance and categorical features. The developed models can be integrated into intrusion detection systems to enhance early threat detection and mitigation. This research contributes to the fields of network security, data analytics, artificial intelligence, and cyber defense strategies.