ACNN-LSTM Based DDoS Attack Detection and Classification

In 2023, globally reported DDoS attacks at the network layer spiked with mitigation of 8.7 million and 85% increase compared to the previous years’ making robust detection strategies vital. As Software Defined Networking (SDN) evolves more quickly to address network management needs, vulnerabilities to DDoS attacks have grown significantly prompting the demand for better detection strategies. It is difficult to detect DDoS attacks characterized by Low-Rate and High-Rate (LR-HR) traits because of their elaborate traffic behavior. DDoS detection is precise in distinguishing between legitimate and malicious traffic hence sparing servers unnecessary exposure to countless attacks hence wastage of resources. From the perspective of avoiding abuse of server resources and preventing the application from a long time of crash in the SDN scenarios, this research contributes to the sustainable use of computational resources more to some extent. This research analyzes three different models: CNN-LSTM, CNN-BiLSTM and CNN-LSTM with Attention Mechanism to enhance the identification of attacks from the LR-HR DDoS 2024 Dataset. The temporal aspects of network traffic are highlighted by LSTM and BiLSTM layers. Evaluation of the CNN-LSTM with Attention mechanism achieves an accuracy of 99.97%, highlighting its dominance over alternative solutions in the identification of DDoS attacks in SDN networks.