Testing Robustness Against Unforeseen Adversaries — HN submission

Mon, 09 Sep 2019 14:25:24 GMT

Testing Robustness Against Unforeseen Adversaries

3 points; 0 discussion replies on HN. Paper and source history. A submission is not an endorsement.

Testing Robustness Against Unforeseen Adversaries — mentioned by mrkoot

Fri, 23 Aug 2019 19:48:48 GMT

For better or worse, this helps establish neural networks that are not (or less?) vulnerable to deception via e.g. fake eyeglasses [ LINK: https://dl.acm.org/citation.cfm?id=2978392 ] and adversarial stickers [ LINK: https://arxiv.org/abs/1712.09665 ]. From the post: "We’ve developed a method to assess whether a neural network classifier can reliably defend against adversarial attacks not seen during training [such as Elastic, Fog, Gabor, and Snow]. Our method yields a new metric, UAR (Unforeseen Attack Robustness), which evaluates the robustness of a single model against an unanticipated…

Paper and discussion history. A mention is not an endorsement.

Testing Robustness Against Unforeseen Adversaries · HackerPapers discussions

Testing Robustness Against Unforeseen Adversaries — HN submission

Testing Robustness Against Unforeseen Adversaries — mentioned by mrkoot