A Stacking Ensemble for Network Intrusion Detection Using Heterogeneous Datasets

Table 6

Multiclass classification results obtained using the UNSW NB-15 dataset.

Index	A	B	D	E	F	G	N	R	S	W	Recall (%)
Analysis	58	0	61	317	32	0	54	1	0	0	11
Backdoor	0	49	79	286	31	1	5	2	1	0	10.79
DOS	3	5	838	2354	66	13	41	11	12	0	25
Exploits	6	6	752	7622	187	33	169	160	25	7	85
Fuzzers	0	2	93	528	2936	7	1217	6	26	0	60.97
Generic	0	2	33	133	14	11512	10	1	2	1	98.32
Normal	19	0	38	163	1260	7	17075	16	16	1	91.82
Reconnaissance	0	2	112	556	5	1	17	2077	2	2	74.8
Shellcode	0	4	6	42	26	0	37	17	184	0	58.22
Worms	0	0	2	18	0	5	0	0	0	15	37.5
Precision (%)	67.44	70	41.6	63.41	64.42	99.42	91.67	90.65	68.65	57.69