Skip to main content

Table 5 Proposed system based on WELM performance on the UNSW dataset

From: Multi-layer intrusion detection system with ExtraTrees feature selection, extreme learning machine ensemble, and softmax aggregation

Attacks

Accuracy with ELM

Accuracy with WELM

Normal

91.26

93.54

Generic

98.16

98.23

Exploits

89.13

90.12

Fuzzers

91.30

91.47

DoS

94.75

94.90

Reconnaissance

94.60

95.33

Analysis

98.96

99.26

Backdoor

99.11

99.11

Shellcode

99.40

99.40

Worms

99.92

99.92

Average

95.66

96.12

Softmax

98.24

98.69