Distributed Denial of Service Attacks Classification and Mitigation

Abstract

Distributed Denial of Service (DDoS) attack introduces new threats and vulnerabilities to the network security, in addition to the ones that already existed. The DDoS attack type is one of the most violent attack type in current years, causing chaos on the entire network system. Most of the research is limited to the classification of DDoS attacks by Machine learning algorithms. Deep learning is not involved in the classification of DDoS attacks and the classes on which classification is applied are not enough. Another challenge is detecting and mitigating DDoS attacks which are not effective and lead to different errors and degraded the accuracy. The existing solutions have adopted old datasets whereas the new DDoS attacks have changed their patterns and types. The proposed model aims to classifying the different classes of DDoS attacks on the new dataset by using deep learning models. The proposed classification classifies the newly released dataset of CICDDOS 2019. The dataset contains complete and current DDoS attack types. The evaluation of the proposed model showed that Long Short Term Memory (LSTM) which is a modification of Recurrent Neural Network (RNN) for the classification of DDoS attacks gives the highest accuracy as compared to SVM (Support Vector Machine). The proposed model achieves better results in terms of accuracy and false positive rate.