Table 5 A summary of the performance of the proposed subword DL models: SW-CNN and SW-LSTM and their character counterparts
From: Use of subword tokenization for domain generation algorithm classification
| Â | Average precision | Average F1 | Average recall |
|---|---|---|---|
Proposed SW-CNN | |||
Word-looking DGA (average of 11 classes) | 0.9355 | 0.9364 | 0.9391 |
Improvement char-CNN (Ren et al. 2020) | 9.59% | 9.70% | 9.89% |
Random-looking DGA (average of 39 classes) | 0.7059 | 0.6728 | 0.6931 |
Improvement over char-CNN (Ren et al. 2020) | 1.70% | 1.89% | 0.64% |
Proposed SW-LSTM | Â | Â | Â |
Word-looking DGA (average of 11 classes) | 0.9473 | 0.9436 | 0.9409 |
Improvement char-LSTM (Cucchiarelli et al. 2021) | 13.64% | 13.07% | 11.65% |
Random-looking DGA (average of 39 classes) | 0.7031 | 0.6731 | 0.6921 |
Improvement char-LSTM (Cucchiarelli et al. 2021) | 0.59% | 1.40% | 0.52% |