3 ms·One pixel attack for fooling deep neural networks3 points by chr1 9y agosharemywin 9y agowonder if gans could be used to help with adversarial attacks. build an adversarial attack generator and train your DNN as the discriminator