Image and Model Transformation with Secret Key for Vision Transformer

In this paper, we propose a combined use of transformed images and vision transformer (ViT) models with secret key. We show for the first time that trained plain can be directly to encrypted on basis ViT architecture, performance is same as when using test addition, proposed scheme does not require any specially prepared data training or network modification, so it also allows us easily update an experiment, effectiveness evaluated in terms degradation model protection image classification task CIFAR-10 dataset.