Adversarial Attack Based on Prediction-Correction

Chen Wan, Fangjun Huang

Deep neural networks (DNNs) are vulnerable to adversarial examples obtained by adding small perturbations to original examples. The added perturbations in existing attacks are main…