Countering Adversarial Images Using Input Transformations [pdf]

arxiv.org

11 pointsstablemap2 comments

How does your work compare to https://openreview.net/pdf?id=S18Su--CW, it seems you only evaluate your model against non-iterative adversaries rather than iterative and stronger ones like PGD.

You might get an answer from Laurens at his post:

https://www.facebook.com/lvdmaaten/posts/10156844136621038

Reply on news.ycombinator.com