Improving the Performance of Machine Learning-Based Network Intrusion Detection Systems on the UNSW-NB15 Dataset

Table 4

Comparing the proposed system with related works on TPR.

Method
Class	GV-SVM [7]	CFS-ABC-AFS [10]	BMM-ADS [11]	5-DNN [13]	RFE-SMOTE [14]	SOM-GA [15]	DO-IDS [16]	The proposed system
Normal	98.47	92.8	93.4	92.8	100	88.3	96.7	97.41
Analysis	—	80.11	83.4	0	18	58.8	6.1	98.89
Backdoors	—	63.4	63.8	34.4	11	64.7	40.3	98.11
DoS	91.22	83.3	89.6	97.7	32	66.9	46.1	82.47
Exploits	67.31	63.7	79.4	1.3	82	79.1	66.3	86.05
Fuzzers	94.39	60.3	52.8	0	89	57.5	38.1	95.08
Generic	96.69	87.3	86.3	57.1	99	89.1	96.9	97.05
Reconnaissance	87.15	49.3	55.6	1.8	76	78.1	82.0	93.16
Shellcode	100	70.9	48.7	0	88	55.0	78.0	99.86
Worms	—	55.3	47.8	0	16	65.9	79.5	99.91