search

robust-ml / robust-ml.github.io

A community-run reference for state-of-the-art adversarial example defenses.
https://www.robust-ml.org/
Creative Commons Attribution Share Alike 4.0 International
49 stars 7 forks source link

ABS Model #2

Closed lukas-schott closed 5 years ago

lukas-schott commented 5 years ago

Name: Towards the first adversarially robust neural network model on MNIST

Authors: Lukas Schott, Jonas Rauber, Matthias Bethge, Wieland Brendel

Paper: https://arxiv.org/pdf/1805.09190.pdf

Code: https://github.com/bethgelab/AnalysisBySynthesis

Venue: 2018 NeurIPS Security Workshop

Does the code implement the robust-ml API and include pre-trained models: yes

Dataset: MNIST

Threat model: L2 (\epsilon = 1.5)

Natural accuracy: 99%

Claims: 80%

anishathalye commented 5 years ago

This looks great -- thank you for the submission! I'll take a look at it and try to update the website sometime in the next few days (after NeurIPS).

anishathalye commented 5 years ago

Thanks again for the submission -- merged in ef26ecb8dbc53bc41235b1a80d2c8b6972990b2a!