Research Article
Trust in Intrusion Detection Systems: An Investigation of Performance Analysis for Machine Learning and Deep Learning Models
Table 12
The performance of deep learning for WSN dataset using LSTM and GRU with all features.
| Model | Evaluation metric | Cross-validation performance | Testing performance | Normal | Grayhole | Blackhole | Scheduling | Flooding | Normal | Grayhole | Blackhole | Scheduling | Flooding |
| LSTM with one layer | TN | 99.99 | 99.98 | 100 | 99.99 | 100 | 98.71 | 99.90 | 99.35 | 92.90 | 99.92 | FP | 0.01 | 0.02 | 0 | 0.01 | 0 | 1.29 | 0.10 | 0.65 | 7.10 | 0.08 | FN | 0 | 0.05 | 0.26 | 3.19 | 23.33 | 7.05 | 6.64 | 43.77 | 0.63 | 10.66 | Accuracy | 99.99 | 99.98 | 100 | 99.98 | 100 | 98.55 | 99.84 | 97.67 | 98.77 | 99.73 | Precision | 100 | 99.92 | 100 | 96.76 | 95 | 66.47 | 88.95 | 77.93 | 99.28 | 95.25 | Recall | 100 | 99.95 | 99.74 | 96.81 | 76.67 | 92.95 | 93.36 | 56.23 | 99.37 | 89.34 | F-score | 100 | 99.93 | 99.87 | 96.75 | 83.67 | 77.51 | 91.10 | 65.33 | 99.32 | 92.20 |
| LSTM with two layer | TN | 98.65 | 99.86 | 99.41 | 91.73 | 99.89 | 98.66 | 99.86 | 99.50 | 93.73 | 99.89 | FP | 1.35 | 0.14 | 0.59 | 8.27 | 0.11 | 1.34 | 0.14 | 0.50 | 6.27 | 0.11 | FN | 8.52 | 7.36 | 47.46 | 0.59 | 12.21 | 8.32 | 3.50 | 43.66 | 0.50 | 11.20 | Accuracy | 98.46 | 99.79 | 97.58 | 98.7 | 99.67 | 98.47 | 99.83 | 97.82 | 98.96 | 99.69 | Precision | 65.15 | 85.19 | 78.39 | 99.16 | 93.3 | 65.31 | 86.19 | 82.01 | 99.36 | 93.53 | Recall | 91.48 | 92.64 | 52.54 | 99.41 | 87.79 | 91.68 | 96.50 | 56.34 | 99.50 | 88.80 | F-score | 76.07 | 88.74 | 62.87 | 99.29 | 90.44 | 76.28 | 91.05 | 66.80 | 99.43 | 91.10 |
| GRU with one layer | TN | 98.67 | 99.87 | 99.3 | 91.82 | 99.88 | 98.78 | 99.89 | 99.24 | 93.29 | 99.93 | FP | 1.33 | 0.13 | 0.7 | 8.18 | 0.12 | 1.22 | 0.11 | 0.76 | 6.71 | 0.07 | FN | 9.61 | 9.42 | 46.84 | 0.65 | 11.89 | 12.86 | 6.76 | 40.94 | 0.57 | 10.66 | Accuracy | 98.45 | 99.79 | 97.5 | 98.66 | 99.68 | 98.47 | 99.83 | 97.68 | 98.86 | 99.75 | Precision | 65.32 | 86.5 | 75.67 | 99.17 | 93.17 | 66.27 | 88.43 | 75.93 | 99.32 | 96.11 | Recall | 90.39 | 90.58 | 53.16 | 99.35 | 88.11 | 87.14 | 93.24 | 59.06 | 99.43 | 89.34 | F-score | 75.7 | 88.47 | 62.23 | 99.26 | 90.56 | 75.29 | 90.77 | 66.44 | 99.37 | 92.60 |
| GRU with two layer | TN | 98.63 | 99.85 | 99.46 | 94.05 | 99.9 | 98.79 | 99.87 | 99.31 | 96.45 | 99.91 | FP | 1.37 | 0.15 | 0.54 | 5.95 | 0.1 | 1.21 | 0.13 | 0.69 | 3.55 | 0.09 | FN | 8.08 | 3.75 | 43.34 | 0.56 | 11.31 | 15.21 | 2.78 | 34.06 | 0.48 | 10.36 | Accuracy | 98.45 | 99.82 | 97.79 | 98.95 | 99.7 | 98.41 | 99.85 | 98.01 | 99.23 | 99.73 | Precision | 65.02 | 85.52 | 80.9 | 99.4 | 94.16 | 65.88 | 86.93 | 79.48 | 99.64 | 94.96 | Recall | 91.92 | 96.25 | 56.66 | 99.44 | 88.69 | 84.79 | 97.22 | 65.94 | 99.52 | 89.64 | F-score | 76.11 | 90.54 | 66.43 | 99.42 | 91.34 | 74.15 | 91.79 | 72.08 | 99.58 | 92.22 |
|
|