Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Kenansville DFT attack #916

Merged
merged 5 commits into from
Nov 17, 2020
Merged

Kenansville DFT attack #916

merged 5 commits into from
Nov 17, 2020

Conversation

yusong-tan
Copy link
Contributor

  • Implements a blackbox signal processing attack called Kenansville
  • Fixes art_experimental MP3 compression
  • Added baseline evaluations configs for this attack

Copy link
Contributor

@kmerchant31892 kmerchant31892 left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I added a few comments to address before this is ready to merge in.

Also, when running the provided configurations with num-eval-batches=10, I noticed that the WER on adversarial examples was actually higher for the defended version than the baseline model. Do we know why this happens? Does this change if I evaluate for more batches?

@kmerchant31892 kmerchant31892 merged commit ab92e27 into twosixlabs:master Nov 17, 2020
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

2 participants