In this paper, we propose a combined use of transformed images and vision transformer (ViT) models transformed with a secret key. We show for the first time that models trained with plain images can be directly transformed to models trained with encrypted images on the basis of the ViT architecture, and the performance of the transformed models is the same as models trained with plain images when using test images encrypted with the key. In addition, the proposed scheme does not require any specially prepared data for training models or network modification, so it also allows us to easily update the secret key. In an experiment, the effectiveness of the proposed scheme is evaluated in terms of performance degradation and model protection performance in an image classification task on the CIFAR-10 dataset.
Hitoshi KIYA
Tokyo Metropolitan University
Ryota IIJIMA
Tokyo Metropolitan University
Aprilpyone MAUNGMAUNG
Tokyo Metropolitan University
Yuma KINOSHITA
Tokyo Metropolitan University
The copyright of the original papers published on this site belongs to IEICE. Unauthorized use of the original or translated papers is prohibited. See IEICE Provisions on Copyright for details.
Copy
Hitoshi KIYA, Ryota IIJIMA, Aprilpyone MAUNGMAUNG, Yuma KINOSHITA, "Image and Model Transformation with Secret Key for Vision Transformer" in IEICE TRANSACTIONS on Information,
vol. E106-D, no. 1, pp. 2-11, January 2023, doi: 10.1587/transinf.2022MUI0001.
Abstract: In this paper, we propose a combined use of transformed images and vision transformer (ViT) models transformed with a secret key. We show for the first time that models trained with plain images can be directly transformed to models trained with encrypted images on the basis of the ViT architecture, and the performance of the transformed models is the same as models trained with plain images when using test images encrypted with the key. In addition, the proposed scheme does not require any specially prepared data for training models or network modification, so it also allows us to easily update the secret key. In an experiment, the effectiveness of the proposed scheme is evaluated in terms of performance degradation and model protection performance in an image classification task on the CIFAR-10 dataset.
URL: https://global.ieice.org/en_transactions/information/10.1587/transinf.2022MUI0001/_p
Copy
@ARTICLE{e106-d_1_2,
author={Hitoshi KIYA, Ryota IIJIMA, Aprilpyone MAUNGMAUNG, Yuma KINOSHITA, },
journal={IEICE TRANSACTIONS on Information},
title={Image and Model Transformation with Secret Key for Vision Transformer},
year={2023},
volume={E106-D},
number={1},
pages={2-11},
abstract={In this paper, we propose a combined use of transformed images and vision transformer (ViT) models transformed with a secret key. We show for the first time that models trained with plain images can be directly transformed to models trained with encrypted images on the basis of the ViT architecture, and the performance of the transformed models is the same as models trained with plain images when using test images encrypted with the key. In addition, the proposed scheme does not require any specially prepared data for training models or network modification, so it also allows us to easily update the secret key. In an experiment, the effectiveness of the proposed scheme is evaluated in terms of performance degradation and model protection performance in an image classification task on the CIFAR-10 dataset.},
keywords={},
doi={10.1587/transinf.2022MUI0001},
ISSN={1745-1361},
month={January},}
Copy
TY - JOUR
TI - Image and Model Transformation with Secret Key for Vision Transformer
T2 - IEICE TRANSACTIONS on Information
SP - 2
EP - 11
AU - Hitoshi KIYA
AU - Ryota IIJIMA
AU - Aprilpyone MAUNGMAUNG
AU - Yuma KINOSHITA
PY - 2023
DO - 10.1587/transinf.2022MUI0001
JO - IEICE TRANSACTIONS on Information
SN - 1745-1361
VL - E106-D
IS - 1
JA - IEICE TRANSACTIONS on Information
Y1 - January 2023
AB - In this paper, we propose a combined use of transformed images and vision transformer (ViT) models transformed with a secret key. We show for the first time that models trained with plain images can be directly transformed to models trained with encrypted images on the basis of the ViT architecture, and the performance of the transformed models is the same as models trained with plain images when using test images encrypted with the key. In addition, the proposed scheme does not require any specially prepared data for training models or network modification, so it also allows us to easily update the secret key. In an experiment, the effectiveness of the proposed scheme is evaluated in terms of performance degradation and model protection performance in an image classification task on the CIFAR-10 dataset.
ER -