Need a candidate who has experience in ML, Deep learning, Tensorflow, Keras -- 2

Defense against adversarial attack: There are several attacks possible on deep learning classifiers(for example, Lets say there is a handwritten image of 5 the deep learning classifier may classify it as 6 or any other number based on attack). There are few defense methods proposed against these attacks like (bit depth reduction, Defense GAN, PixelCNN, Image quilting, Randomized diversification) Here you have to work on the defense against those attacks like modifying existing defense methods. I need a new method which is modification of Defense GAN. Like replacing with different classifier, activation functions or anything which should be different from Defense GAN