From what little I have read on the topic adversarial attacks are succesfull on neural networks because in order to train them succesfully we design them to ope

From what little I have read on the topic adversarial attacks are succesfull on neural networks because in order to train them succesfully we design them to operate on a linear regime as much as possible. This linearity compounded with the high dimensionality of the problem (images) is what makes easy to find small perturbations that make the output oscillate wildly. My point is that "attacking machine learning" maybe should be renamed "attacking neural networks"