Table 7 A summary of the performance of the proposed integrated schemes XG + SW-CNN and XG + SW-LSTM
From: Use of subword tokenization for domain generation algorithm classification
| Â | Average precision | Average F1 | Average recall |
|---|---|---|---|
Proposed XG + SW-CNN | |||
Overall | 0.8016 | 0.7950 | 0.7997 |
Improvement over XGBoost | 5.13% | 4.98% | 5.07% |
Improvement over SW-CNN | 5.62% | 8.47% | 6.88% |
Improvement over CNN-BiLSTM (Cucchiarelli et al. 2021) | 5.52% | 6.31% | 5.27% |
Random-looking DGAs | 0.7641 | 0.7554 | 0.7605 |
Improvement over XGBoost | − 0.25% | − 0.54% | − 0.51% |
Improvement over SW-CNN | 8.24% | 12.28 | 9.72% |
Improvement over CNN-BiLSTM (Cucchiarelli et al. 2021) | 6.29% | 7.51% | 6.08% |
Word-looking DGAs | 0.9300 | 0.9327 | 0.9391 |
Improvement over XGBoost | 25.07% | 24.48% | 24.45% |
Improvement over SW-CNN | − 0.40% | − 0.30% | 0% |
Improvement over CNN-BiLSTM (Cucchiarelli et al. 2021) | 3.58% | 3.52% | 3.70% |
Proposed XG + SW-LSTM | |||
Overall | 0.8023 | 0.7970 | 0.8010 |
Improvement over XGBoost | 5.22% | 5.24% | 5.24% |
Improvement over SW-LSTM | 6.01% | 8.58% | 6.98% |
Improvement over CNN-BiLSTM (Cucchiarelli et al. 2021) | 5.61% | 6.57% | 5.44% |
Random-looking DGAs | 0.7636 | 0.7559 | 0.7603 |
Improvement over XGBoost | − 0.31% | − 0.47% | − 0.54% |
Improvement over SW-LSTM | 8.60% | 12.30% | 9.85% |
Improvement over CNN-BiLSTM (Cucchiarelli et al. 2021) | 6.22% | 7.59% | 6.05% |
Word-looking DGAs | 0.9427 | 0.9418 | 0.9409 |
Improve over XGBoost | 26.54% | 25.57% | 24.69% |
Improvement over SW-LSTM | − 0.49% | − 0.19% | 0% |
Improvement over CNN-BiLSTM (Cucchiarelli et al. 2021) | 4.99% | 4.53% | 3.90% |