Research Article
An Intrusion Detection Method Based on Fully Connected Recurrent Neural Network
Table 1
Model detection accuracy (2-class) under different structures and Learningrates.
| | | KDDTrain+ | KDDTest+ | KDDTest−21(%) |
| HiddenNodes = 20,
Learningrate = 0.01 | 99.40% | 79.37% | 60.76 |
| HiddenNodes = 20,
Learningrate = 0.1 rate = 0.1 | 99.79% | 83.18% | 68.23 |
| HiddenNodes = 20,
Learningrate = 0.5 | 99.81% | 83.09% | 67.84 |
| HiddenNodes = 60,
Learningrate = 0.01 | 99.39% | 78.72% | 59.54 |
| HiddenNodes = 60,
Learningrate = 0.1 | 99.79% | 81.06% | 64.08 |
| HiddenNodes = 60,
Learningrate = 0.5 | 99.87% | 83.11% | 67.82 |
| HiddenNodes = 80,
Learningrate = 0.01 | 99.29% | 79.16% | 60.34 |
| HiddenNodes = 80,
Learningrate = 0.1 | 99.81% | 83.28% | 68.55 |
| HiddenNodes = 80,
Learningrate = 0.5 | 99.85% | 82.66% | 66.99 |
| HiddenNodes = 120,
Learningrate = 0.01 | 99.28% | 78.55% | 59.25 |
| HiddenNodes = 120,
Learningrate = 0.1 | 99.79% | 82.48% | 66.83 |
| HiddenNodes = 120,
Learningrate = 0.5 | 99.87% | 80.97% | 63.69 |
| HiddenNodes = 240,
Learningrate = 0.01 | 99.69% | 80.69% | 63.28 |
| HiddenNodes = 240,
Learningrate = 0.1 | 99.69% | 80.67% | 63.28 |
| HiddenNodes = 240,
Learningrate = 0.5 | 99.87% | 80.97% | 63.69 |
|
|
