LA-GRU: Building Combined Intrusion Detection Model Based on Imbalanced Learning and Gated Recurrent Unit Neural Network

Table 6

Comparison of overall detection performance among different IDMs.

Method		ACC(%)	DR(%)	FAR(%)	Dataset	Training Dataset Size
Imbalanced Learning	CANN+SMOTE[9]	98.99	99.56	0.557	NSL-KDD	125,973
	MHCVF[10]	98.04	95.57	1.38	KDD CUP 99	494,021
	DENDRON[11]	97.55	95.97	1.08	NSL-KDD	125,973
	I-NGSA[12]	99.37	99.24	N/A	NSL-KDD	125,973
Shallow Learning	SVM[2]	94.22	92.99	3.46	KDD CUP 99	145,585
	OS-ELM[3]	98.66	99.01	1.74	NSL-KDD	125,973
	TLMD[4]	93.32	93.11	0.761	KDD CUP 99	86,000
	GA-LR[35]	99.90	99.81	0.105	KDD CUP 99	494,021
Deep Learning	CNN+LSTM[13]	99.68	97.78	0.07	KDD CUP 99	2,466,929
	S-NADE[14]	97.85	97.85	2.15	KDD CUP 99	494,021
	DNN[15]	99.20	99.27	0.85	NSL-KDD	125,973
	SCDNN[16]	92.03	92.23	7.90	NSL-KDD	62,986
Proposed Method	LA-GRU	99.04	98.92	0.134	NSL-KDD	73,906