Skip to main content

Table 2 Comparison of campaign stages classification results

From: TriCTI: an actionable cyber threat intelligence discovery system via trigger-enhanced neural network

Model Original training set Augmentation training set
Accuracy Precision Recall F1 Accuracy Precision Recall F1
BiLSTM 81.23 81.45 81.39 81.27 81.27 81.54 81.55 81.38
BiLSTM-Att 80.76 81.31 80.99 80.94 82.02 82.38 82.17 82.10
BERT 82.99 83.77 82.93 82.87 83.63 84.15 83.68 83.65
TriCTI 86.38 86.55 86.57 86.43 86.99 87.16 87.10 87.02
  1. We run the model on the original labeled training set and the augmented training set. Accuracy, Precision, Recall, and F1 score (macro) are the average value over 3 runs