Optimizing DDoS Traffic Classification Using Enhanced Deep Neural Network Techniques

Abstract: This study investigates the effectiveness of an enhanced Deep Neural Network (DNN) model for classifying Distributed Denial of Service (DDoS) traffic. The proposed model integrates an Adaptive Attention Layer (AAL) technique to improve detection accuracy. Experiments were conducted using a full and a reduced feature sets with three dataset sizes, namely; 4K, 40K, and 225K samples. Results reveal a positive correlation between the number of features, dataset size, and model performance metrics. The study highlights the pivotal role of the AAL in dynamically prioritizing the most relevant features, enabling the model to more effectively detect subtle anomalies in high-dimensional traffic data. This attention mechanism significantly reduces both false positives and false negatives. Across all configurations, the proposed model consistently achieved high performance and near-perfect accuracy, precision, recall, and F1-Scores. These results underscore the model’s robustness, adaptability, and potential for real-world deployment in modern DDoS detection systems. Keywords: DDoS Classification, Deep Neural Network, Data Normalization.