Attacking Machine Learning with Adversarial Examples

https://mobile.twitter.com/TheGregYang/status/1139205510475833346

Are Accuracy and Robustness Correlated? We find that adversarial examples are mostly transferable across similar network topologies, and we demonstrate that better machine learning models are less vulnerable to adversarial examples.

EXPLAINING AND HARNESSING ADVERSARIAL EXAMPLES

Deep Neural Networks are Easily Fooled

Numerical accuracy may help against adversarial attacks: https://arxiv.org/abs/1704.01547

https://www.youtube.com/watch?v=hDlHpBBGaKs

https://mobile.twitter.com/aleks_madry/status/1125430774767988736