Skip to main content

Table 8 Comparison between models with adversarial training and models without adversarial training

From: On the combination of data augmentation method and gated convolution model for building effective and robust intrusion detection

  Model ACCorig ACCadv Perturbation rate Crafting rate
Adversarial sequences attack CNN 96.40% 34.40% 19.50% 64.20%
  GCNN 97.50% 35.40% 22.70% 63.70%
Adversarial training CNN 96.90% 45.20% 28.00% 53.20%
  GCNN 97.90% 60.70% 28.60% 37.90%