Project - Defense against adversarial attack
dataset: MNIST, CIFAR10
adversarial attacks : There are several attacks possible on deep learning classifiers(for example, Lets say there is a handwritten image of 5 the deep learning classifier(here I used CNN) may classify it as 6 or any other number based on attack). github : [login to view URL]
Defense GAN: [login to view URL], [login to view URL]
Implement this defense GAN and make some changes like using different classifier, activation functions or anything which should be different from Defense GAN.
Write the report on proposed methodology