Combination of Hybrid Feature Selection and LSTM-AE Neural Network for Enhancing DDOS Detection in SDN

Distributed Denial of Service (DDoS) attacks pose significant threats in Software-Defined Network (SDN) environments. To enhance DDoS detection in SDN, this study presents a novel approach that combines Hybrid Feature Selection and Long Short-Term Memory (LSTM)-Autoencoder (AE) neural network. The feature selection process initially utilizes Information Gain (IG) to select 50% of the most important ones. Subsequently, the SHapley Additive exPlanations (SHAP) method is employed to identify relevant and interdependent features. The LSTM-AE neural network then captures the temporal characteristics and nonlinear patterns in the system's response, creating a low-dimensional data representation. Experimental evaluation using an SDN dataset demonstrates the effectiveness of the proposed approach, achieving an overall accuracy of 99.23%. The hybrid feature selection and LSTM-AE model offer improved DDoS detection capabilities in SDN environments.