Enhanced DDoS Attack Detection Using Advanced Deep Learning Techniques

This paper presents a rigorous comparative analysis of three distinct studies dedicated to the detection of distributed denial-of-service (DDoS) attacks. Leveraging the comprehensive CICDDoS2019 dataset, which encompasses a rich variety of both DDoS and normal traffic instances, as well as diverse attack patterns, these studies meticulously explore an array of detection methodologies. The first study employs Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU), and Recurrent Neural Network (RNN) models, while the second study focuses on Deep Convolutional Autoencoder (DCAE) and basic Autoencoder (AE) models. The third study introduces innovative hybrid architectures, amalgamating LSTM, Convolutional Neural Network (CNN), and Deep Neural Network (DNN) approaches, alongside a novel CNN model. The results of our analysis under-score the GRU model’s exceptional performance, achieving an outstanding accuracy of $\mathbf{9 9. 5 4 \%}$ with notably superior execution time. This finding unequivocally emphasizes the effectiveness of deep learning methodologies in enhancing the resilience of cyber defenses against the relentless evolution of modern threats.