Intelligent Detection of DDoS Attacks in Software Defined Networks Using Deep Learning

Many security issues arise due to the increasing prevalence and magnitude of DDoS attacks. This is particularly the case for Software Defined Networks that have a certain degree of management centralization as this exacerbates their weaknesses. The present research is focused on the use of deep learning techniques for the purpose of recognizing DDoS attacks in SDN systems. We inspected the performance of eight separate deep learning models including; CNN, DNN, RNN, LSTM, MLP and combined CNN+DNN model and the DataSet_SDN. The dataset contains normal instances as well as attack instances which have over 80 features including packet number and protocol type. To achieve better results from the models, data preprocessing processes such as one-hot encoding and feature selection were performed. Accuracy, precision, recall and F1 Score are the metrics that were used to evaluate the proposed model. In this case, the underwriting of the DNN+CNN hybrid model is better than any DNN based algorithm, which achieved an accuracy of 99.90% and F1 Score of 99.46. These results have great significance because they imply that SDN networks do have the potential to mitigate DDoS attacks through the deep learning consideration.