A data-driven network intrusion detection system using feature selection and deep learning.

Network intrusion detection system (NIDS) is an important line of defense for network security as network attacks become more frequent. In this paper, we propose a data-driven NIDS based on feature selection and deep learning, named FS-DL. FS-DL focuses on improving data quality, using methods such as standard deviation and association rule mining to remove a large number of redundant features, reduce computational load, and improve detection accuracy. To balance detection accuracy and time cost, FS-DL uses a simple neural network structure with only three layers and minimizes the number of neurons as much as possible. Experimental results show that FS-DL only requires a small number of traffic features to obtain better detection performance. In addition, we have designed an NIDS based on FS-DL, deployed in the software-defined networking (SDN) controller for online detection of abnormal traffic. [ABSTRACT FROM AUTHOR]