Evasion Attack in Show and Tell Model

Recently, deep learning technology has been applied to various fields with high performance and various services. Image recognition is also used in various fields with high performance by incorporating deep learning technology. However, deep learning technology is vulnerable to evasion attacks that cause the model to be misclassified by modulating the original image. In this paper, we generate an adversarial example using the forward-backward-splitting iterative procedure. Then perform an evasion attack on the show and tell model.