From: TIM: threat context-enhanced TTP intelligence mining on unstructured threat data
Text level | Model | Precision | Recall | F1 |
---|---|---|---|---|
Document-level | 0.437 | 0.500 | 0.608 | |
Li et al. (2019) | 0.444 | 0.509 | 0.547 | |
Sentence-level | Doc2Vec + linear SVC | 0.859 | 0.881 | 0.870 |
Doc2Vec + DT | 0.853 | 0.857 | 0.855 | |
Doc2Vec + RF | 0.895 | 0.902 | 0.899 | |
Bi-LSTM + Attention (GloVe) | 0.871 | 0.923 | 0.896 | |
TextCNN (GloVe) | 0.913 | 0.914 | 0.927 | |
fastText | 0.936 | 0.895 | 0.915 | |
BERT-CLS | 0.940 | 0.935 | 0.935 | |
TCENet | 0.970 | 0.973 | 0.971 |