Research on Network Traffic Intrusion Detection Based on CFS-S Feature Selection Algorithm

  • Qiao Ding, Shizhuang Yin, Lijun Liu, Chao Wang


With the rapid development of big data, cloud computing, cloud storage and other technologies, network bandwidth in the 5G era keeps increasing, resulting in an exponential growth trend of network traffic. In order to ensure the security of information, network traffic intrusion detection system must acquire and analyze the traffic data dynamically. However, in the acquired traffic data features, considerable parts of the features are redundant. Therefore, reducing and selecting the appropriate features can improve the speed of intrusion detection system. In this research, by improving the CFS algorithm, a hybrid feature selection strategy based on the CFS-S algorithm is proposed, which fully integrates the advantages of the traditional feature selection methods, such as short time consuming filtering method and high precision of wrapping method. The policy was applied to the publicly available NSL_KDD data set, 13 features were successfully selected from the original 41. The experimental results show that compared with the traditional feature selection methods, the feature selection strategy based on CFS-S algorithm can greatly shorten the detection time without significantly reducing the detection rate.