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

Table 6

Comparing the proposed system with related works on F1-score.

Method
Class	RFE-SMOTE [14]	DO-IDS [16]	CNN-BiLSTM [18]	The proposed system
Normal	100	93.0	84.99	98.43
Analysis	28	5.3	9.69	99.44
Backdoors	18	21.9	8.97	99.05
DoS	34	39.9	29.55	90.39
Exploits	72	70.8	67.89	88.45
Fuzzers	91	54.2	37.47	96.34
Generic	99	98.3	98.85	98.41
Reconnaissance	84	85.3	62.54	93.11
Shellcode	87	48.6	30.95	99.92
Worms	25	78.7	10.75	97.34