3 ms·Universal adversarial perturbations, fooling deep-learning classifiers1 points by lazugod 10y ago