LNNLS-KH: A Feature Selection Method for Network Intrusion Detection

As an important part of intrusion detection, feature selection plays a significant role in improving the performance detection. Krill herd (KH) algorithm is efficient swarm intelligence with excellent data mining. To solve problem low efficiency and high false positive rate detection caused by increasing high-dimensional data, improved krill based on linear nearest neighbor lasso step (LNNLS-KH) proposed for network The number selected features classification accuracy are introduced into fitness evaluation function LNNLS-KH algorithm, physical diffusion motion individuals transformed nonlinear method. Meanwhile, optimization performed updated position order to derive global optimal solution. Experiments show that retains 7 NSL-KDD dataset 10.2 CICIDS2017 average, which effectively eliminates redundant while ensuring accuracy. Compared CMPSO, ACO, KH, IKH algorithms, it reduces 44%, 42.86%, 34.88%, 24.32% dataset, 57.85%, 52.34%, 27.14%, 25% respectively. increased 10.03% 5.39%, 8.63% 5.45%. Time decreased 12.41% 4.03% average. Furthermore, quickly jumps out local solution shows good iteration curve, convergence speed,