TransDDoS: Transformer-Based Model for Intelligent Detection of DDoS Attacks

Abstract

Distributed Denial of Service (DDoS) attacks pose a critical threat to the stability and availability of modern network infrastructures. This study proposes the application of a pure Transformer architecture for the effective detection of DDoS attacks, utilizing the CICDDoS2019 dataset. The research workflow includes data preprocessing (cleaning, normalization, and one-hot encoding), feature selection using the Random Forest algorithm, data splitting (80% training and 20% testing), model training, and performance evaluation. The results show that the Transformer model achieves an accuracy of 99.82%, precision of 99.80%, recall of 99.83%, and F1-score of 99.82%. This approach outperforms previous methods such as CNN, Deep Neural Networks, Deep Q-Networks, and ensemble models, which typically reach accuracy levels between 90% and 99%. The superior performance of the Transformer is attributed to its self-attention mechanism, which effectively captures complex patterns in network traffic data. The key contributions of this study include the novel implementation of a Transformer model in the field of network intrusion detection, the integration of RF feature selection to enhance model efficiency, and a comprehensive empirical evaluation demonstrating improved results over traditional approaches. The findings indicate that the Transformer is a highly promising approach for developing intelligent, early-warning systems to counter large-scale cyberattacks. Future work may explore real-time deployment and adaptive learning capabilities to respond to emerging threats.