A GAN and Feature Selection-Based Oversampling Technique for Intrusion Detection

Table 8

Performance of different methods on the CICIDS-2017 dataset.

		SMOTE	ADASYN	K-SMOTE	G-SMOTE	ACGAN-SVM	GAN	Proposed
NB	Acc	0.6652	0.6816	0.7238	0.7543	0.7263	0.7227	0.7239
	F1	0.4854	0.2653	0.3105	0.4425	0.4836	0.4797	0.4806
DT	Acc	0.9393	0.9585	0.9459	0.2490	0.9746	0.9721	0.9770
	F1	0.8220	0.8895	0.8454	0.3414	0.9331	0.9264	0.9397
RF	Acc	0.9969	0.9950	0.9684	0.5847	0.9692	0.9703	0.9984
	F1	0.9921	0.9872	0.9130	0.4856	0.9154	0.9187	0.9960
GBDT	Acc	0.9921	0.9947	0.9928	0.9779	0.9888	0.9719	0.9962
	F1	0.9801	0.9867	0.9817	0.9468	0.9710	0.9333	0.9902
K-NN	Acc	0.9839	0.9823	0.9850	0.9694	0.9847	0.9830	0.9819
	F1	0.9599	0.9562	0.9622	0.9267	0.9615	0.9613	0.9549
ANN	Acc	0.8161	0.3498	0.8925	0.5665	0.9294	0.9227	0.9667
	F1	0.6811	0.3774	0.7153	0.4334	0.8185	0.8311	0.9189