CoChat

Testing human ability to detect ‘deepfake’ images of human faces

Open access
Bray, Sergi D, Johnson, Shane D, Kleinberg, Bennett · 2023 · Journal of Cybersecurity

The DOI matches its Crossref record; 98 of the 98 references it cites were found in a registry or an index.

Abstract

‘Deepfakes’ are computationally created entities that falsely represent reality. They can take image, video, and audio modalities, and pose a threat to many areas of systems and societies, comprising a topic of interest to various aspects of cybersecurity and cybersafety. In 2020, a workshop consulting AI experts from academia, policing, government, the private sector, and state security agencies ranked deepfakes as the most serious AI threat. These experts noted that since fake material can propagate through many uncontrolled routes, changes in citizen behaviour may be the only effective defence. This study aims to assess human ability to identify image deepfakes of human faces (these being uncurated output from the StyleGAN2 algorithm as trained on the FFHQ dataset) from a pool of non-deepfake images (these being random selection of images from the FFHQ dataset), and to assess the effectiveness of some simple interventions intended to improve detection accuracy. Using an online survey, participants (N = 280) were randomly allocated to one of four groups: a control group, and three assistance interventions. Each participant was shown a sequence of 20 images randomly selected from a pool of 50 deepfake images of human faces and 50 images of real human faces. Participants were asked whether each image was AI-generated or not, to report their confidence, and to describe the reasoning behind each response. Overall detection accuracy was only just above chance and none of the interventions significantly improved this. Of equal concern was the fact that participants’ confidence in their answers was high and unrelated to accuracy. Assessing the results on a per-image basis reveals that participants consistently found certain images easy to label correctly and certain images difficult, but reported similarly high confidence regardless of the image. Thus, although participant accuracy was 62% overall, this accuracy across images ranged quite evenly between 85 and 30%, with an accuracy of below 50% for one in every five images. We interpret the findings as suggesting that there is a need for an urgent call to action to address this threat.

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 · 10.1093/cybsec/tyad011

Cites

98 of 98 references verified0 citation errors · 0 not checkable

Statistics

Figures

Figure 2 Example of highlighting explanatory regions of image during experiment.
Figure 2 Example of highlighting explanatory regions of image during experiment.CC-BY · From the PDF
Figure 3 Use of 'tell-tale signs' in participants' reasoning behind response decisions, by decision correctness (where participants' responses contained 'Deepfake' as the selected label).
Figure 3 Use of 'tell-tale signs' in participants' reasoning behind response decisions, by decision correctness (where participants' responses contained 'Deepfake' as the selected label).CC-BY · From the PDF
Image 3
CC-BY · From the PDF
Image 4
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 98 references

Research works printed but not linked: 37

  1. Merriam Webster, 2020, Words we’re watching: “Deepfake.”
  2. Conor, 2021, Prescriptivist vs descriptivist—what’s the difference?, One Minute English
  3. Foster Wallace, 2006, Authority and american usage, 66
  4. Deepfake
  5. 2023, Audio deepfake, Wikipedia
  6. Dash, 2023, Are ChatGPT and deepfake algorithms endangering the cybersecurity industry? A review, Int J Appl Sci Eng, 10, 1
  7. 2022, Neural style transfer, Wikipedia
  8. Style Transfer Guide
  9. DALL·E 2, OpenAI
  10. Damiani, 2019, A Voice Deepfake Was Used To Scam A CEO Out Of $243,000, Forbes
  11. 2020, Snapshot Paper—Deepfakes and Audiovisual Disinformation, GOVUK
  12. 2019, The Hypnotic Allure of the AI Art Generator, Sothebys.com
  13. Alfo Media, 2018, On the Edge—AI Generated Rock Music Composed by AIVA
  14. VICE News, 2018, How Does A.I. Art Stack Up Against Human Art? (HBO)
  15. 2020, Inside the Score, Can Artificial Intelligence Make Good Music?
  16. Galvani, 2021, You Shall Be, SoundCloud
  17. Oest, 2020, Sunrise to Sunset: Analyzing the End-to-end Life Cycle and Effectiveness of Phishing Attacks at Scale
  18. 2017, Stop Unwanted Robocalls and Texts, Federal Communications Commission
  19. 2020, The Draft Online Safety Bill and the legal but harmful debate Date Published: 24 January 2022 UK Parliament, GOVUK
  20. Gow, 2021, The Scary Truth Behind the FBI Warning: Deepfake Fraud is Here And it’s Serious—We Are Not Prepared, Forbes
  21. 2021, The Grandparent Scam, Federal Bureau of Investigation
  22. Albahar, 2019, Deepfakes: threats and Countermeasures Systematic Review, J Theor Appl Inf Technol, 97, 3242
  23. Reuters Institute, 2020, Digital News Report 2020
  24. Reuters Institute, 2018, Digital News Report 2018
  25. 2018, The Strategy and Tactics of the Pro-Kremlin Disinformation Campaign, EU vs Disinfo
  26. Great Britain, Committee on Standards in Public Life, 2017, Intimidation in Public Life: A Review by the Committee on Standards in Public Life
  27. Stevens, 2019, How blockchain could help combat deepfakes, Decrypt
  28. Rashid, 2021, Blockchain technology for combating deepfake and protect video/image integrity, J. Korea Multimed. Soc., 24, 1044
  29. Sato, 2020, What Is Account Takeover Fraud and How Can You Prevent It?, Experian
  30. Guerra, 2010, Exploring the Spam Arms Race to Characterize Spam Evolution
  31. 2020, NVlabs/Ffhq-Dataset
  32. Metz, 2019, These people do not exist. Why websites are churning out fake images of people (and cats), CNN
  33. O'Sullivan, 2019, Now fake Facebook accounts are using fake faces, CNN
  34. O'Sullivan, 2020, How fake faces are being weaponized online, CNN
  35. Vincent, 2019, A spy reportedly used an AI-generated profile picture to connect with sources on LinkedIn, The Verge
  36. 2021, Deepest-Project/MelNet
  37. 2021, This MP Does Not Exist, VOLE.wtf

Cited by

See all 152

Metadata

Authors
Bray, S. D., Johnson, S. D., & Kleinberg, B.
Affiliations
  • University College London (GB)
  • Tilburg University (NL)
Corresponding author
Bray, Sergi D
Venue
Journal of Cybersecurity
Published
2023-01-01
Volume
9
Issue
1
ISSN
  • 2057-2085 (ISSN-L)
  • 2057-2093
OpenAlex
W4381851764
Cited by
96
References
95
Citations by year
2026: 27 · 2025: 34 · 2024: 30 · 2023: 5
FWCI
8.73
Type
journal article
Language
en
Open access
Locations
  • Journal of Cybersecurity (publishedVersion, cc-by)
  • Research portal (Tilburg University) (submittedVersion, cc-by)
Funders
EPSRC (EP/S022503/1)
Topics
  • Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Digital Media Forensic Detection
  • Social Sciences › Sociology and Political Science › Misinformation and Its Impacts
  • Life Sciences › Neuroscience › Cognitive Neuroscience › Face Recognition and Perception
Keywords
deepfakes, deepfake detection, face images, StyleGAN2, confidence calibration, misinformation, AI risks, cybersecurity

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: Testing human ability to detect ‘deepfake’ images of human faces. https://app.cochat.ai/papers/10.1093/cybsec/tyad011
The paper
Bray, S. D., Johnson, S. D., & Kleinberg, B. (2023). Testing human ability to detect ‘deepfake’ images of human faces. Journal of Cybersecurity. https://doi.org/10.1093/cybsec/tyad011

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.