How to generate a defended dataset to evaluate DF and RF?

[DUCK-lnbby]

My approach is as follows:
Process the CW dataset by labeling each traffic record with a domain name ID and number. Each file contains detailed information about the packets, including timestamps and directions.
After processing with extract-list.py, cluster.py, refinement.py, and regularization.py, generate the defended traffic. At this point, the filenames remain the same as the original undisturbed traffic, and then they are converted back into npz files for DF and RF attacks. However, my RF results consistently show 1.05%, and I am unsure which details need to be corrected. If possible, I would greatly appreciate any guidance.