Sharing Wanet poisoned weights

[soroushhashemifar]

Hi there. I checked Wanet weights but I think it's the weights after cleaning. Could you please share the poisoned weights of Wanet attack? Thank you so much

[soroushhashemifar]

And the same problem with weights of SIG attack. The initial accuracy on the poisoned data-set is around 0.0

[soroushhashemifar]

Also the checkpoint provided for Dynamic attack is clean. Please provide the poisoned model if it is possible.

[bboylyg]

Hi,
Thank you for your interest in our work. The issue at hand is most likely caused by a mismatch of poisoned test data. Regrettably, the test data is too large to be uploaded. I strongly recommend that you utilize the script located at this URL [https://github.com/bboylyg/RNP/blob/main/train_backdoor.py] to replicate these attacks. The script includes code for both Wavnet and Dynamic attacks.