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

Table 5

Binary-class classification performance on KDD 99 in terms of TPR (%), FPR (%), and Acc (%). Note that EFS-DNN outperforms all baselines with the small number of features.

Method	TPR	FPR	Acc	# Features
LSSVM-IDS + FMIFS [9]	99.46	0.13	99.79	19
DBN + ensemble SVM [18]	99.82	n/a	n/a	All
SwiftIDS [12]	99.76	n/a	n/a	All
PIO [10]	97.40	9.71	94.39	10
SVM-IDS [24]	97.62	n/a	98.95	20
EFS-DNN	99.91	0.09	99.92	14