Student or educator? Sign up with your school email and get 6 months of Pro free.
CoChat

The DOI matches its Crossref record; 48 of the 52 references it cites were found in a registry or an index.

Pixel-Based Image Encryption Without Key Management for Privacy-Preserving Deep Neural Networks

Open access

Abstract

We present a novel privacy-preserving scheme for deep neural networks (DNNs) that enables us not to only apply images without visual information to DNNs but to also consider the use of independent encryption keys for both training and testing images for the first time. In this paper, a novel pixel-based image encryption method that maintains important features of original images is proposed for privacy-preserving DNNs. For training, a DNN model is trained with images encrypted by using the proposed method with independent encryption keys. For testing, the model enables us to apply both encrypted images and plain images for image classification. Therefore, there is no need to manage keys. In addition, the proposed method allows us to perform data augmentation in the encrypted domain. In an experiment, the proposed method is applied to well-known networks, that is, deep residual networks and densely connected convolutional networks, for image classification. The experimental results demonstrate that the proposed method, under the use of independent encryption keys, can maintain a high classification performance, and it is robust against ciphertext-only attacks (COAs). Moreover, the results confirm that the proposed scheme is able to classify plain images as well as encrypted images, even when data augmentation is carried out in the encrypted domain.

Overview

The claims and statistics of this paper haven’t been checked yet.

Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.

A few free checks a day · papers already verified open instantly · results are public

Integrity signals

Registry

Crossref · 10.1109/access.2019.2959017

Cites

48 of 48 references verified0 citation errors · 0 not checkable

Statistics

Figures

FIGURE 1 Examples of images. (a) Original image ( U V D 96 96 ). (b) Image encrypted by block-based encryption [11], [12] (Block size D 4 4 ). (c) Image encrypted by block-based encryption [18] (Block size D 4 4 ). (d) Image encrypted by pixel-based image encryption [20].
FIGURE 1 Examples of images. (a) Original image ( U V D 96 96 ). (b) Image encrypted by block-based encryption [11], [12] (Block size D 4 4 ). (c) Image encrypted by block-based encryption [18] (Block size D 4 4 ). (d) Image encrypted by pixel-based image encryption [20].CC-BY · From the PDF
FIGURE 2 Frameworks of model training for image classification. (a) A DNN model is trained by Enc ( f d ( T )) . (b) f d ( Enc ( T )) is used for training a model.
FIGURE 2 Frameworks of model training for image classification. (a) A DNN model is trained by Enc ( f d ( T )) . (b) f d ( Enc ( T )) is used for training a model.CC-BY · From the PDF
FIGURE 3 Frameworks of model testing for image classification. (a) XQ D Enc ( Q ) is applied to trained model. (b) Trained model is tested by XQ D Q .
FIGURE 3 Frameworks of model testing for image classification. (a) XQ D Enc ( Q ) is applied to trained model. (b) Trained model is tested by XQ D Q .CC-BY · From the PDF
FIGURE 4 Proposed image encryption.
FIGURE 4 Proposed image encryption.CC-BY · From the PDF
FIGURE 5 Examples of images encrypted by proposed encryption method, where Fig. 1(a) is original. (a) Image encrypted by NP. (b) Image encrypted by NP and CS.
FIGURE 5 Examples of images encrypted by proposed encryption method, where Fig. 1(a) is original. (a) Image encrypted by NP. (b) Image encrypted by NP and CS.CC-BY · From the PDF
FIGURE 7 Network architecture of reconstruction model. Each box denotes multi-channel feature map produced by each layer. Number of channels is denoted above each box. Feature map resolutions are U V . Kernel size and stride of locally connected layers are (1,1).
FIGURE 7 Network architecture of reconstruction model. Each box denotes multi-channel feature map produced by each layer. Number of channels is denoted above each box. Feature map resolutions are U V . Kernel size and stride of locally connected layers are (1,1).CC-BY · From the PDF
Image 7
CC-BY · From the PDF
FIGURE 9 64 filters with size of 3 3 in first convolution layer of ResNet18. Data augmentation was carried out in plain domain. (a) Model trained by using plain images. (b) Model trained by using encrypted images with NP under the use of different keys. (c) Model trained by using encrypted images with NP and CS under the use of different keys.
FIGURE 9 64 filters with size of 3 3 in first convolution layer of ResNet18. Data augmentation was carried out in plain domain. (a) Model trained by using plain images. (b) Model trained by using encrypted images with NP under the use of different keys. (c) Model trained by using encrypted images with NP and CS under the use of different keys.CC-BY · From the PDF
FIGURE 10 Image classification accuracy when testing DNN models (ResNet-18) with plain images Q , where models were trained with encrypted images. Horizontal axis corresponds to conditions of images used for training DNNs. Note that same key and different keys correspond to the encryption key conditions used for encrypting training and testing images.
FIGURE 10 Image classification accuracy when testing DNN models (ResNet-18) with plain images Q , where models were trained with encrypted images. Horizontal axis corresponds to conditions of images used for training DNNs. Note that same key and different keys correspond to the encryption key conditions used for encrypting training and testing images.CC-BY · From the PDF
Image 10
CC-BY · From the PDF
Image 11
CC-BY · From the PDF
FIGURE 11 Image classification accuracy when testing DNN models (ResNet-18) with Enc ( Q ) , where models were trained with following data augmentation conditions: (a) Without data augmentation, (b) Shifting, and (c) Horizontal flipping. Note that same key and different keys correspond to the encryption key conditions used for encrypting training and testing images.
FIGURE 11 Image classification accuracy when testing DNN models (ResNet-18) with Enc ( Q ) , where models were trained with following data augmentation conditions: (a) Without data augmentation, (b) Shifting, and (c) Horizontal flipping. Note that same key and different keys correspond to the encryption key conditions used for encrypting training and testing images.CC-BY · From the PDF
FIGURE 12 Image classification accuracy when testing DNN models (DenseNet) with Enc ( Q ) , where depth and growth rate of DenseNet were 40 and 12, respectively. Note that same key and different keys correspond to the encryption key conditions used for encrypting training and testing images.
FIGURE 12 Image classification accuracy when testing DNN models (DenseNet) with Enc ( Q ) , where depth and growth rate of DenseNet were 40 and 12, respectively. Note that same key and different keys correspond to the encryption key conditions used for encrypting training and testing images.CC-BY · From the PDF
Image 14
CC-BY · From the PDF
Image 15
CC-BY · From the PDF
(b) NP with different keys
(b) NP with different keysCC-BY · From the PDF
(d) NP and CS with different keys
(d) NP and CS with different keysCC-BY · From the PDF
Image 18
CC-BY · From the PDF
Image 19
CC-BY · From the PDF
ter Student Award, in 2018.
ter Student Award, in 2018.CC-BY · From the PDF

Verification

Claims the paper makes

The claims and statistics of this paper haven’t been checked yet.

Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.

A few free checks a day · papers already verified open instantly · results are public

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

The self-check needs the paper’s full text.

Citations

Cites

Show all 52 references

Research works printed but not linked: 1

  1. szegedy, 2013, Intriguing properties of neural networks, arXiv 1312 6199

Cited by

See all 111

Metadata

Authors
Sirichotedumrong, W., Kinoshita, Y., & Kiya, H.
Affiliations
Tokyo Metropolitan University (JP)
Venue
IEEE Access
Published
2019
Volume
7
Pages
177844-177855
ISSN
2169-3536 (ISSN-L)
OpenAlex
W2995164854
Cited by
105
References
52
Citations by year
2026: 4 · 2025: 12 · 2024: 17 · 2023: 16 · 2022: 21 · 2021: 19 · 2020: 16
FWCI
6.26
Type
journal article
Language
en
Open access
Locations
  • IEEE Access (publishedVersion, cc-by)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
Funders
Japan Society for the Promotion of Science (17H03267)
Topics
  • Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Chaos-based Image/Signal Encryption
  • Physical Sciences › Computer Science › Computer Networks and Communications › Cryptography and Data Security
  • Physical Sciences › Computer Science › Artificial Intelligence › Cryptographic Implementations and Security
Keywords
key management, ciphertext-only attack, ResNet, DenseNet, image privacy

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

How to cite

Cite CoChat for the checks on this page, and the paper for its own findings.

This page
CoChat. (2026). Citation check: Pixel-Based Image Encryption Without Key Management for Privacy-Preserving Deep Neural Networks. https://app.cochat.ai/papers/10.1109/access.2019.2959017
The paper
Sirichotedumrong, W., Kinoshita, Y., & Kiya, H. (2019). Pixel-Based Image Encryption Without Key Management for Privacy-Preserving Deep Neural Networks. IEEE Access. https://doi.org/10.1109/access.2019.2959017

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.