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

Table 5

Comparing the proposed system with related works on FPR.

Method
Class	GV-SVM [7]	RF [8]	5-DNN [13]	DO-IDS [16]	LCC-MI-SVM-FS [17]	RF [19]	The proposed system
Normal	0.04	—	0.285	0.033	0.383	—	0.0074
Analysis	—	0.0056	0	0.39	0	0.016	0
Backdoors	—	0.0005	0.013	0.597	0	0.018	0
DoS	0.08	0.002	0	0.539	0.0018	0.01	0
Exploits	0.06	0.014	0	0.337	0.1081	0.009	0.0317
Fuzzers	0.01	—	0	0.619	0.0169	0.012	0.0040
Generic	0.01	0.00091	0.166	0.031	0.0234	0.008	0.0021
Reconnaissance	0.02	0.007	0.008	0.180	0	0.014	0.0070
Shellcode	0.09	0.006	0	0.220	0	0.016	0.0001
Worms	—	0	0	0.205	0	0.019	0.0536