CoChat
Sign Up
Cited by 7Advanced Steganography and Watermarking TechniquesDigital Media Forensic DetectionChaos-based Image/Signal Encryption

The DOI matches its Crossref record; 2 of the 2 references it cites were found in a registry or an index.

An Analysis of PhotoDNA

Open access
Steinebach, Martin·2023·Proceedings of the 18th International Conference on Availability, Reliability and Security
DOI 10.1145/3600160.3605048
Open this paper in CoChat — free accountOpen PDF

Abstract

PhotoDNA is a popular hash used to combat CSAM. So far, only limited information has been provided about this hash in terms of its performance. In this paper, we provide an overview of robustness and false positives, as well as some basic observations about its structure. We show that it is robust against typical image processing such as lossy compression. On the other hand, it is only of limited robustness against cropping. We also give some suggestions for improving the algorithm or its use.

Overview

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

Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.

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

Integrity signals

Registry

Crossref · 10.1145/3600160.3605048

Cites

2 of 2 references verified0 citation errors · 0 not checkable

Statistics

Figures

Figure 1 Sections of PhotoDNA. The numbers indicate the positions of the first 12 6x6 quadrants. The quadrants overlap by 2 pixels. The red quadrant 6 and the blue quadrant 8 illustrate the horizontal and vertical overlap
Figure 1 Sections of PhotoDNA. The numbers indicate the positions of the first 12 6x6 quadrants. The quadrants overlap by 2 pixels. The red quadrant 6 and the blue quadrant 8 illustrate the horizontal and vertical overlapCC-BY · From the PDF
Figure 2 Hash example of one image of the coco dataset.Top: original image, second line: gray scale and resize to 26x26 pixel, third line: assignment to quadrants, bottom: values in the of the hash of the first 12 quadrants
Figure 2 Hash example of one image of the coco dataset.Top: original image, second line: gray scale and resize to 26x26 pixel, third line: assignment to quadrants, bottom: values in the of the hash of the first 12 quadrantsCC-BY · From the PDF
Figure 3 Mean of 144 hash elements
Figure 3 Mean of 144 hash elementsCC-BY · From the PDF
Figure 4 Illustration of hash distribution
Figure 4 Illustration of hash distributionCC-BY · From the PDF
Figure 5 Mean of 144 hash elements sorted by direction
Figure 5 Mean of 144 hash elements sorted by directionCC-BY · From the PDF
Figure 6 Three test images with simple structures
Figure 6 Three test images with simple structuresCC-BY · From the PDF
Table 4: Robustness in percent for 2,000 test images against attacks
Table 4: Robustness in percent for 2,000 test images against attacksCC-BY · From the PDF
Image 8
CC-BY · From the PDF
Figure 8 Top 10 low distances, 600 randomly selected images, 160.000 images in hash data base
Figure 8 Top 10 low distances, 600 randomly selected images, 160.000 images in hash data baseCC-BY · From the PDF
Figure 9 Example of image pair with small distance of 221
Figure 9 Example of image pair with small distance of 221CC-BY · From the PDF

Verification

Claims the paper makes

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

Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.

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 2 references

Research works printed but not linked: 21

  1. 10.1145/3407023.3409212 · DOI
  2. Olena Buchko 2022. Classification of Confidential Images Using Neural Hash. 5 (2022), 68–71.
  3. 10.5120/10765-5578 · DOI
  4. 10.23919/MIPRO.2017.7973569 · DOI · Paper page
  5. 10.1016/j.image.2019.115713 · DOI
  6. 10.21300/19.3.2018.593 · DOI
  7. Tsung-Yi Lin, Michael Maire, Serge J. Belongie, Lubomir D. Bourdev, Ross B. Girshick, James Hays, Pietro Perona, Deva Ramanan, Piotr Dollár, and C. Lawrence Zitnick. 2014. Microsoft COCO: Common Objects in Context. CoRR abs/1405.0312 (2014). arXiv:1405.0312http://arxiv.org/abs/1405.0312 · arXiv
  8. Dat Tien Nguyen Firoj Alam Ferda Ofli and Muhammad Imran. [n.d.]. Automatic Image Filtering on Social Networks Using Deep Learning and Perceptual Hashing During Crises. http://arxiv.org/pdf/1704.02602v1 · arXiv
  9. 10.1007/978-3-642-35515-8_15 · DOI · Paper page
  10. 10.1145/3538969.3544461 · DOI
  11. 10.1145/3339252.3341487 · DOI
  12. Martin Steinebach, Huajian Liu, and York Yannikos. 2012. Forbild: Efficient robust image hashing. In Media Watermarking, Security, and Forensics 2012, Vol. 8303. SPIE, 195–202.
  13. 10.1109/ARES.2014.85 · DOI
  14. 10.1007/978-3-319-14289-0_8 · DOI
  15. 10.1007/s11042-012-1188-8 · DOI
  16. 10.1109/TKDE.2018.2837745 · DOI · Paper page
  17. 10.1016/j.ijleo.2014.05.015 · DOI
  18. 10.1016/j.aeue.2013.02.009 · DOI
  19. Shaharyar Ahmed Khan Tareen and Zahra Saleem. 2018. A comparative analysis of sift, surf, kaze, akaze, orb, and brisk. In 2018 International conference on computing, mathematics and engineering technologies (iCoMET). IEEE, 1–10.
  20. 10.5555/1193214.1194040 · DOI
  21. 10.1117/12.876617 · DOI

Cited by

Sorted by citations

Metadata

Authors
Steinebach, M.
Affiliations
Fraunhofer Institute for Secure Information Technology (DE)
Corresponding author
Steinebach, Martin
ORCID
Steinebach, Martin
Venue
Proceedings of the 18th International Conference on Availability, Reliability and Security
Publisher
ACM
Published
2023-08-29
Pages
1-8
DOI
10.1145/3600160.3605048
OpenAlex
W4385688380
Cited by
7
References
2
Citations by year
2026: 3 · 2025: 3 · 2024: 1
FWCI
0.75
Type
conference paper
Language
en
Open access
Licence
Locations
  • International Conference on Availability, Reliability and Security (ARES) (publishedVersion, cc-by)
  • Fraunhofer-Publica (Fraunhofer-Gesellschaft) (submittedVersion)
Funders
Bundesministerium für Bildung und Forschung
Topics
  • Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Advanced Steganography and Watermarking Techniques
  • Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Digital Media Forensic Detection
  • Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Chaos-based Image/Signal Encryption
Keywords
content safety, child pornography, image hashing, hash robustness, false positives, lossy image compression, image cropping, image processing

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