Ensemble-Learning Approach to DDoS-Attack Detection using Stacking, Meta-Learning, and Adversarial Training
Open AccessAs the Internet continues to grow every day with the exponential use of Internet of Things (IoT) devices, the introduction of 5th generation Internet (5G), and other emerging technologies, the distributed denial of service (DDoS) attacks affecting multi-billion-dollar businesses have also become prevalent and complex. This trend poses significant challenges for network security. To address these challenges, this research introduces a novel Adversarial Weighted Stacked Model (AWSM), a significant extension of the stacking model, integrating adversarial training and dynamic ensemble weighting to improve the accuracy of DDoS attack characterization or detection.The study began with an evaluation of several standard machine learning models, such as: Random Forest (RF), Decision Tree (DT), K-Nearest Neighbors (KNN), eXtreme Gradient Boosting (XGB), Bagging Classifier (BC), Support Vector Machine (SVM), and Naive Bayes (NB). The individual models demonstrated strong performance on test data, with accuracies ranging from 88.43% to 99.87%. An advanced ensemble learning approach, Stacking, was then employed to further improve the detection capabilities, achieving a competitive accuracy of 99.87% on the test data. The research also focused on the generalization of the models by testing them on unseen data. The unseen data results showcased the models' robustness, with accuracies ranging from 88.27% to 99.87%. Notably, the stacking model and the AWSM exhibited excellent generalization capabilities, maintaining high accuracy rates of 99.88% and 99.91%, respectively, highlighting their potential for real-world application. These findings underscore the effectiveness of ensemble learning approaches, particularly the stacking model and the novel AWSM, in enhancing DDoS attack detection. Moreover, incorporating adversarial training bolstered the models' resilience against evasion tactics, improving their overall performance and robustness. The research represents a significant contribution to the academic discourse on DDoS attack detection, offering the ability to better understand the effectiveness of advanced ensemble learning techniques in real-world scenarios. The outcomes accentuate the possibilities engraved in ensemble learning methods, such as Stacking and AWSM, for fortifying network security against the escalating threat of DDoS attacks. These findings pave the way for future research into more resilient and adaptable defense mechanisms against the rapidly evolving landscape of cyber threats.
- All rights reserved
Notice to Authors
If you are the author of this work and you have any questions about the information on this page, please use the Contact form to get in touch with us.