Attacking machine learning with adversarial examples

Re:Attacking machine learning with adversarial examples by aidenn0 on Thursday February 16, 2017 @06:44PM (#13664688)

Without knowing much about ML, it seems that using two (or more) very different methods could be a reasonable defense; if the methods are sufficiently different then it will get exponentially harder to find a gradient that fools all the methods; what to do when the outputs strongly disagree is a good question, but switching to a failsafe mode seems better than what we have now.

[ Parent | on HN ]

Re:Attacking machine learning with adversarial examples by lerid on Friday February 17, 2017 @03:25AM (#13666782)

These adversarial inputs have been shown to generalize to separately trained models.

[ Parent | on HN ]

Re:Attacking machine learning with adversarial examples by aidenn0 on Saturday February 18, 2017 @09:12PM (#13677840)

I was thinking use the same training set but different ML techniques; It's been almost 20 years since I took an AI class, but RNNs weren't the only thing in there...

[ Parent | on HN ]