Enhanced Network Defense: Optimized Multi-Layer Ensemble for DDoS Attack Detection

In today's digitally connected world, Distributed Denial of Service (DDoS) attacks remain a formidable challenge, undermining the stability of network infrastructures and demanding robust detection strategies. This research explores advanced methodologies for DDoS detection by conducting a comparative analysis of machine learning and deep learning approaches using the CICDDoS2019 dataset. Initially, a hybrid machine learning framework is implemented, integrating K-Means clustering for pre-labeling the dataset and employing supervised models such as Random Forest (RF), Extreme Gradient Boosting (XGBoost), Adaptive Boosting (AdaBoost), Support Vector Machine (SVM), and Artificial Neural Network (ANN). This approach achieves an accuracy of 99.46%, showcasing its effectiveness while highlighting challenges like manual feature selection and limited scalability for complex datasets. A novel hybrid deep learning architecture is proposed to overcome these challenges, combining Convolutional Neural Networks (CNN) for spatial feature extraction and Long Short-Term Memory (LSTM) networks for temporal sequence learning. This automated feature extraction mechanism eliminates reliance on manual intervention, ensuring adaptability to evolving attack patterns. The proposed CNN-LSTM model demonstrates an impressive accuracy of 99.84%, significantly outperforming traditional machine learning models. Additionally, the model's adaptability and resilience against dynamic attack behaviours position it as a reliable solution for real-time DDoS mitigation. This study emphasizes the growing relevance of deep learning techniques in enhancing cyber security and underscores the potential of hybrid architectures in effectively detecting and mitigating modern cyber threats. The findings provide valuable insights into developing scalable, high-performance systems capable of addressing the ever-evolving nature of DDoS attacks.