RS4

Reading Summary: Detection and defense of DDoS attack-based on deep learning in OpenFlow-based SDN

Chloe Li, CSC 466

Mar 15, 2021

Technologies are getting more complicated from time to time, the traditional methods for DDoS (Distributed denial of service) attack detection and defense technology has the problem of poor adaptability, low detection efficiency, error or leakage report etc.

It is also difficult to detect the attacks by using the general machine learning methods with the complexity of intrusion data.

The problem is important at the time of paper publication. Before they design this SDN architecture, the SDN network security was one of the key research issues. DDoS was the largest threat for organizations. This SDN architecture was verified in real-time network environment for defending and detecting DDoS attacks.

Since deep learning has strong learning ability, it is possible to apply it to DDoS attack detection. Based on deep learning, the paper proposed a DDoS attack detection and defense method. By using the ISCX data set, the verification accuracy of the DDoS attack is as high as 98% or 99% in the model training phase. This SDN architecture improves the accuracy of DDoS attack detection. It also reduces the degree of dependence on the software and hardware environment.

When the DDoS attack detection result generates the OpenFlow flow entries, it can effectively clean the DDoS attack traffic to reduce DDoS attacks in SDN networks. The SDN architecture has high detection accuracy and has little dependence on hardware and software devices. This architecture also simplifies the difficulty of upgrading the DDoS attack detection strategy.

The rule base of the DDoS attack is difficult to identify and update. Since the DDoS attack rule base matching increase the processing time of switches, the transmission delay of forwarding the normal network data packets is raised. The processing delay of the entire network environment will be added.

I may use different data set to test the model to see if the results are similar or different.

Reference:

Li C, Wu Y, Yuan X, et al. Detection and defense of DDoS attack–based on deep learning in OpenFlow‐based SDN. Int J Commun Syst. 2018;31:e3497. https://doi.org/10.1002/dac.3497