The Defense of Adversarial Example with Conditional Generative Adversarial Networks

<div>The prediction accuracy of the defense mechanism for the MNIST dataset (the range of epoch is from 2 to 20): the experimental results of the target models (a) Anet and (b) Bnet.</div>

Security and Communication Networks

fig5

Figure 5

Figure 5: The Defense of Adversarial Example with Conditional Generative Adversarial Networks 