EFS-DNN: An Ensemble Feature Selection-Based Deep Learning Approach to Network Intrusion Detection System

Table 8

Five-class classification results on KDD 99 in terms of Acc (%). EFS-DNN outperforms all baselines.

Method	Normal	DoS	Probe	R2L	U2R
Flexible neural tree [44]	99.19	98.75	98.39	99.09	99.70
Radial SVM [45]	n/a	98.94	97.11	97.78	97.80
IPSO-RBF [46]	n/a	93.95	95.80	96.02	95.20
LSSVM-IDS + FMIFS [9]	99.79	99.86	99.91	99.92	99.97
KNN + PSO [20]	99.57	99.91	94.41	99.73	99.77
SVM-IDS [24]	98.95	98.99	95.36	98.76	99.01
EFS-DNN	99.82	99.96	99.95	99.90	99.97