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

Table 6

Performance of different methods on the NSL-KDD dataset.

		SMOTE	ADASYN	K-SMOTE	G-SMOTE	ACGAN-SVM	GAN	Proposed
NB	Acc	0.8475	0.8090	0.8246	0.8027	0.7595	0.7771	0.8052
	F1	0.8552	0.8132	0.8307	0.7973	0.7395	0.7600	0.7993
DT	Acc	0.7856	0.7583	0.7736	0.8101	0.7983	0.7646	0.8092
	F1	0.7746	0.7392	0.7624	0.8047	0.7946	0.7564	0.8451
RF	Acc	0.7571	0.7494	0.7837	0.7822	0.7571	0.7618	0.8293
	F1	0.7363	0.7249	0.7715	0.7694	0.7365	0.7424	0.8635
GBDT	Acc	0.7800	0.7780	0.7840	0.7749	0.7741	0.7785	0.8328
	F1	0.7751	0.7646	0.7774	0.7684	0.7686	0.7737	0.8669
SVM	Acc	0.8080	0.8025	0.7821	0.8407	0.7781	0.7798	0.7555
	F1	0.8023	0.7963	0.7686	0.8425	0.7632	0.7654	0.7918
K-NN	Acc	0.7924	0.7890	0.7921	0.7921	0.7758	0.7758	0.7985
	F1	0.7823	0.7812	0.7821	0.7835	0.7599	0.7599	0.7921
ANN	Acc	0.7661	0.7850	0.7802	0.7850	0.7625	0.7587	0.7924
	F1	0.7577	0.7865	0.7759	0.7814	0.7455	0.7377	0.7931