Improved Detection of Four Attacks Based on Particle Method
Tao Yu, Zhen Liu · 2021 17th International Conference on Computational Intelligence and Security (CIS) · 2021
Network security has attracted more and more attention. In order to improve the defense ability against network attacks, researchers have put forward many solutions, and an intrusion detection system is one of them. In this paper, the NSL-KDD data set contains four types of attacks (U2R, R2L, Probe, DoS) and non-attacks. By adding the particle method, the density of each data is calculated and combined with the characteristic quantity of each data, and the data is divided into five categories. In order to improve the computational efficiency and remove the features of low correlation and high interference, we reduce the dimension of data features. The attack detection rates of U2R, R2L, Probe, and DoS, are calculated by machine learning methods such as DT, NN, SVM, K-NN, and NB. Experiments show that the particle method improves the detection rate of four types of attacks.