Synthesizing Robust Adversarial Examples · HackerPapers discussions

Synthesizing Robust Adversarial Examples — mentioned by ben_w

Tue, 03 Sep 2024 22:36:21 GMT

> But powerful humans are also gamed, and usually more easily than computers. It's the "more easily" that I'd dispute. A computer can be examined in detail, its weaknesses probed ruthlessly and systematically until a vulnerability is found. Any one human will have traits in common with others, and differences from those others. For example, your own question: > Haven't you ever told an exec what they wanted to hear, or refrained from telling a manager something that might make you or your project look bad? this would be a dangerous similarity, one that can be manipulated — but not…

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

Synthesizing Robust Adversarial Examples — mentioned by dijksterhuis

Wed, 22 Jul 2020 23:01:54 GMT

Applying some form of transformation is a weak countermeasure to adversarial examples. Especially simple ones such as low pass filters. https://arxiv.org/pdf/1707.07397.pdf Neural Networks do not learn what humans learn. They can learn completely different and sometimes much smaller features.

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

Synthesizing Robust Adversarial Examples — mentioned by fxtentacle

Mon, 03 Feb 2020 10:45:41 GMT

The problem that I see with supervised training of a linear classifier after unsupervised training is that if the unsupervised network is large enough, it allows the supervised trainer to choose the working components. As shown in [1] that can lead to randomly initialized networks working well, too, meaning that this does not necessarily show that the unsupervised training produced useful features. I would instead suggest to train a categorization classifier unsupervised, too, for example using mutual information loss with the correct number of categories, as suggested in [2]. Afterwards, one…

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

Synthesizing Robust Adversarial Examples — mentioned by bnegreve

Mon, 01 Apr 2019 05:41:40 GMT

I'm not sure why hackers prefer to hack large scale computer systems rather than individual humans, but they do. So we have to protect neural networks against adversarial examples for the same reason we have to protect databases against sql injections. > so what's the critical difference here? If neural networks are deployed at scale in self driving cars, a single bug could trigger millions of accidents. Printing an adversarial example on billboards would lead to crashes all around the country. Are we going to assume no one is going to try? (btw: real world adversarial examples are easy to…

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

Synthesizing Robust Adversarial Examples — mentioned by sgillen

Wed, 06 Mar 2019 19:43:01 GMT

I think they are using the wrong word here, I think by "determinism" they mean interpretable. You train your big neural network to classify images, it works fine on the test set but then one day in the wild you discover your network classified a turtle as a rifle [0]. Why did it do that? what can we change to fix that, how could we have seen this coming? Those answers won't come easily compared to a more old school system. [0] https://arxiv.org/abs/1707.07397

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

Synthesizing Robust Adversarial Examples — mentioned by nealeratzlaff

Tue, 02 Jan 2018 22:53:55 GMT

I think for most people that hurdle of understanding is immense, and isn't likely to get any better. ML/DL infrastructure will continue to improve and deploying AI systems will become more accessible as time goes on. But I don't see familiarity with (non) convex optimization, statistical learning, topology, etc. becoming mainstream. Without these its hard to see a reasonable path forward for AI. For me this year in DL was a lot narrower. it mostly revolved around systems becoming more human in capability. I think most of this is in the article already: We can now hopefully say that ImageNet…

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

Synthesizing Robust Adversarial Examples — HN submission

Sun, 05 Nov 2017 07:05:41 GMT

Synthesizing Robust Adversarial Examples

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

Synthesizing Robust Adversarial Examples — mentioned by RodgerTheGreat

Mon, 16 Oct 2017 20:09:49 GMT

I think the author's conclusion- that this scenario is unrealistic and would never happen given today's understanding of machine learning techniques- is extremely optimistic. NNs are demonstrably[1] not robust image classifiers. In my opinion, it's far more dangerous to downplay the limitations of this technology and embolden snake-oil purveyors than it is to demand an inconvenient degree of rigor and caution in reporting results. [1] https://arxiv.org/abs/1707.07397#

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