- 期刊
Research on Network Anomaly Data Flow Intrusion Detection and Defense Under Self-Defending Network Architecture
摘要
A self-defending network (SDN) is easier to control and scale than traditional Internet architectures, but they also have to face malicious attacks, primarily Distributed Denial of Service (DDoS) attacks. This paper briefly introduced the basic architecture of SDN and DDoS attacks in SDN. Convolutional Neural Network (CNN) was combined with Long Short Term Memory (LSTM) for the detection and defense against DDoS attacks. The simulation experiments were conducted on the CNN+LSTM algorithm in the SDN built in the laboratory, and the CNN+LSTM algorithm was compared with K-means and CNN algorithms. The experimental results showed that the CNN+LSTM detection algorithm had higher recognition accuracy and efficiency; the SDN switch had a lower traffic share under the CNN+LSTM detection algorithm than the other two detection algorithms when facing DDoS attacks; as the number of DDoS attacks increased, the CNN+LSTM algorithm maintained stable recognition accuracy, recognition efficiency and switch traffic share.