The DOI matches its DataCite record.
Robust Video Hash Extraction
Open accessAbstract
Publication in the conference proceedings of EUSIPCO, Viena, Austria, 2004
Overview
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
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?
For all five evaluated perceptual hashing algorithms (pHash continuous and discrete, dHash, aHash, and PDQ), the attack avoids detection with modified images very similar to the original.
“We perform a largescale extensive evaluation of five commonly used perceptual hashing algorithms: pHash (continuous and discrete) [12,63], dHash, aHash and PDQ [28] and show that, for all of them, our attack manages avoid detection with modified images very similar to the original.”
From Adversarial Detection Avoidance Attacks: Evaluating the robustness of perceptual hashing-based client-side scanning (2021) · cites this paper as [12]
pHash, aHash, dHash, and Facebook's PDQ are commonly used algorithms for image deduplication or retrieval.
“They are commonly used algorithms for image deduplication or retrieval.”
From Adversarial Detection Avoidance Attacks: Evaluating the robustness of perceptual hashing-based client-side scanning (2021) · cites this paper as [12]
For pHash, k = 32 and the 8x8 submatrix of pixels from rows and columns 2 to 9 is preserved; for PDQ, k = 64 and the 16x16 submatrix from rows and columns 2 to 17 is preserved.
“For the pHash algorithm, k = 32 and the 8 × 8 submatrix of pixels from the intersection of rows and columns 2 to 9 is preserved, while for PDQ k = 64 and the 16 × 16 submatrix of pixels from the intersection of rows and columns 2 to 17 is preserved.”
From Adversarial Detection Avoidance Attacks: Evaluating the robustness of perceptual hashing-based client-side scanning (2021) · cites this paper as [12]
pHash, aHash and dHash output a 64-bit hash while PDQ outputs a 256-bit hash for any input image.
“pHash, aHash and dHash outputs a 64-bit hash while PDQ outputs a 256-bit hash for any input image.”
From Adversarial Detection Avoidance Attacks: Evaluating the robustness of perceptual hashing-based client-side scanning (2021) · cites this paper as [12]
Citations
Cited by
Sorted by citations
Jain, S., Cretu, A.-M., & de Montjoye, Y.-A. (2021) · arXiv · doi:10.48550/arxiv.2106.09820 · Cited by 21
Jiang, N., Sun, B., Kim, N., Sim, T., & Han, J. (2026) · 2026 IEEE Symposium on Security and Privacy Workshops (SPW) · doi:10.1109/spw72489.2026.00029
Metadata
- Authors
- Coskun, B., & Sankur, B.
- Affiliations
- Boğaziçi University (TR)
- ORCID
- Sankur, B.
- Venue
- Zenodo
- Publisher
- Zenodo
- Published
- 2004-09-06
- Pages
- 2295-2298
- OpenAlex
- W1586376507
- Cited by
- 16
- References
- 0
- Citations by year
- 2023: 1 · 2022: 1 · 2019: 1 · 2018: 1 · 2015: 1 · 2013: 1 · 2012: 1
- FWCI
- 0.75
- Type
- conference paper
- Language
- en
- Open access
- green
- doi.org
- Licence
- Locations
- Zenodo (CERN European Organization for Nuclear Research) (cc-by)
- Topics
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Advanced Steganography and Watermarking Techniques
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Video Analysis and Summarization
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Chaos-based Image/Signal Encryption
- Keywords
- video hashing, robustness, video fingerprinting
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
