Research Article

Trust in Intrusion Detection Systems: An Investigation of Performance Analysis for Machine Learning and Deep Learning Models

Table 16

The performance of deep learning for KDD dataset using LSTM and GRU with all features.

ModelEvaluation metricCross-validation performanceTesting performance
DOSNormalProbeR2LU2RDOSNormalProbeR2LU2R

LSTM with one layerTNR98.799.8799.2491.8399.8899.9899.96100.0099.99100.00
FPR1.30.130.768.170.120.020.040.000.010.00
FNR10.429.6446.160.6811.670.010.061.247.6266.67
Accuracy98.4699.7997.4798.6399.6899.9999.9599.9999.9799.99
Precision65.6186.1674.4899.1793.07100.0099.8099.5597.2266.67
Recall89.5890.3653.8499.3288.3399.9999.9498.7692.3833.33
F-score75.688.262.2899.2490.62100.0099.8799.1594.7444.44

LSTM with two layerTNR99.8999.7510099.9510099.9299.9399.9999.96100.00
FPR0.110.2500.0500.080.070.010.040.00
FNR0.180.352.6615.571000.010.292.2210.85100.00
Accuracy99.8399.7399.9799.9199.9999.9899.8999.9899.9399.99
Precision99.9798.8599.5882.2099.9899.7099.3784.920
Recall99.8299.6597.3484.43099.9999.7197.7889.150.00
F-score99.999.2598.4483.27099.9999.7098.5786.980

GRU with one layerTNR99.8899.5610099.9610099.9299.90100.0099.97100.00
FPR0.120.4400.0400.080.100.000.030.00
FNR0.370.323.0714.281000.040.252.9311.73100.00
Accuracy99.6799.5899.9799.9299.9999.9599.8799.9799.9399.99
Precision99.979899.4684.57099.9899.5399.5486.740
Recall99.6399.6896.9385.72099.9699.7597.0788.270.00
F-score99.898.8398.1785.05099.9799.6498.2987.500

GRU with two layerTNR99.8999.5399.9999.9610099.9699.87100.0099.97100.00
FPR0.110.470.010.0400.040.130.000.030.00
FNR0.380.282.6223.671000.050.192.6714.96100.00
Accuracy99.6799.5699.9799.999.9999.9599.8699.9799.9399.99
Precision99.9897.8799.286.37099.9999.4399.6487.350
Recall99.6299.7297.3876.33099.9599.8197.3385.040.00
F-score99.898.7898.2876.73099.9799.6298.4786.180