Research Article
Trust in Intrusion Detection Systems: An Investigation of Performance Analysis for Machine Learning and Deep Learning Models
Table 17
The performance of deep learning for KDD dataset using LSTM and GRU with selected features.
| Model | Evaluation metric | Cross-validation performance | Testing performance | | DOS | Normal | Probe | R2L | U2R | DOS | Normal | Probe | R2L | U2R |
| LSTM with one layer | TNR | 66.96 | 98.75 | 100 | 99.99 | 100 | 78.06 | 98.50 | 100.00 | 99.99 | 100.00 | FPR | 33.04 | 1.25 | 0 | 0.01 | 0 | 21.94 | 1.50 | 0.00 | 0.01 | 0.00 | FNR | 0.95 | 30.63 | 100 | 79.18 | 100 | 1.06 | 19.57 | 100.00 | 77.13 | 100.00 | Accuracy | 92.98 | 93.51 | 99.13 | 99.79 | 99.99 | 94.98 | 95.27 | 99.15 | 99.79 | 99.99 | Precision | 92.91 | 92.35 | 0 | 84.04 | 0.0 | 95.06 | 92.09 | 0.0 | 82.11 | 0.0 | Recall | 99.05 | 69.37 | 0 | 20.82 | 0 | 98.94 | 80.43 | 0.00 | 22.87 | 0.00 | F-score | 95.84 | 77.59 | 0.0 | 39.48 | 0.0 | 96.96 | 85.87 | 0.0 | 35.78 | 0.0 |
| LSTM with two layer | TNR | 81.53 | 98.37 | 99.99 | 100 | 100 | 85.83 | 98.37 | 100.00 | 99.99 | 100.00 | FPR | 18.47 | 1.63 | 0.01 | 0 | 0 | 14.17 | 1.63 | 0.00 | 0.01 | 0.00 | FNR | 1.24 | 15.93 | 87.69 | 94.94 | 100 | 1.40 | 11.50 | 76.09 | 76.83 | 100.00 | Accuracy | 95.49 | 95.82 | 99.22 | 99.76 | 99.99 | 96.17 | 96.60 | 99.35 | 99.79 | 99.99 | Precision | 95.82 | 91.78 | 84.58 | 91.37 | 0.0 | 96.74 | 92.19 | 98.18 | 86.81 | 0.0 | Recall | 98.76 | 84.07 | 12.31 | 5.06 | 0 | 98.60 | 88.50 | 23.91 | 23.17 | 0.00 | F-score | 97.26 | 87.71 | 33.94 | 39.52 | 0.0 | 97.66 | 90.31 | 38.46 | 36.57 | 0.0 |
| GRU with one layer | TNR | 78.73 | 99.22 | 100 | 100 | 100 | 79.00 | 99.19 | 100.00 | 100.00 | 100.00 | FPR | 21.27 | 0.78 | 0 | 0 | 0 | 21.00 | 0.81 | 0.00 | 0.00 | 0.00 | FN | 0.36 | 18.26 | 100 | 100 | 100 | 0.36 | 18.15 | 100.00 | 100.00 | 100.00 | Accuracy | 95.68 | 96.11 | 99.13 | 99.75 | 99.99 | 95.72 | 96.09 | 99.15 | 99.74 | 99.99 | Precision | 95.25 | 95.77 | 0.0 | 0.0 | 0.0 | 95.29 | 95.66 | 0.0 | 0.0 | 0.0 | Recall | 99.64 | 81.74 | 0 | 0 | 0 | 99.64 | 81.85 | 0.00 | 0.00 | 0.00 | F-score | 97.4 | 88.2 | 0.0 | 0.0 | 0.0 | 97.42 | 88.22 | 0.0 | 0.0 | 0.0 |
| GRU with two layer | TNR | 82.9 | 98.94 | 99.99 | 99.98 | 100 | 83.25 | 99.34 | 99.98 | 99.99 | 100.00 | FPR | 17.1 | 1.06 | 0.01 | 0.02 | 0 | 16.75 | 0.66 | 0.02 | 0.01 | 0.00 | FNR | 0.77 | 14.4 | 86.83 | 73.34 | 100 | 0.46 | 14.11 | 76.09 | 76.54 | 100.00 | Accuracy | 96.14 | 96.56 | 99.24 | 99.8 | 99.99 | 96.45 | 96.93 | 99.34 | 99.79 | 99.99 | Precision | 96.13 | 94.62 | 91.67 | 80.4 | 0.0 | 96.20 | 96.59 | 93.08 | 83.33 | 0.0 | Recall | 99.23 | 85.6 | 13.17 | 26.66 | 0 | 99.54 | 85.89 | 23.91 | 23.46 | 0.00 | F-score | 97.65 | 89.88 | 35.28 | 39.92 | 0.0 | 97.84 | 90.92 | 38.05 | 36.61 | 0.0 |
|
|