Collecting False Positives/Negatives of RNN model

[llmhyy]

Hi Yilong @Yilong94 ,

Would you please first sample some false positive/negative of the RNN model based on the training traffics? More specifically, we need to collect the following traffic on both training and testing/validation set:
(1) the normal traffic predicted with low accuracy (<0.5) by normal model;
(2) the normal traffic predicted with high accuracy (>0.8) by attack model;
(3) the attack traffic predicted with high accuracy (>0.8) by normal model;
(4) the attack traffic predicted with low accuracy (<0.5) by attack model.

We may need to use the tool to check those traffic and think about the solution for exsiting problem. Thanks!