Neural networks are known to be vulnerable to adversarial examples. In this note, we evaluate the two white-box defenses that appeared at CVPR 2018 and find they are ineffective: when applying existing techniques, we can reduce the accuracy of the defended models to 0%.
翻译:众所周知,神经网络容易受到对抗性例子的影响。 在本说明中,我们评估了CVPR 2018上出现的两种白箱防御,发现它们无效:在应用现有技术时,我们可以将防御模型的准确性降低到0%。