Pith. sign in

REVIEW 4 major objections 6 minor 32 references

Challenges in GenAI and Authentication: a scoping review

T0 review · 4 major / 6 minor · reviewed 2026-08-06 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read A structured scoping review distills the GenAI-authentication literature to 13 core studies and maps its challenges, threats, and gaps.

desk verdict A useful annotated list of 13 papers on GenAI and authentication, but the screening arithmetic does not add up, so the central claim of a comprehensive synthesis is unsupported. read the letter →

arxiv 2507.11775 v1 pith:EAHTBEKX submitted 2025-07-15 cs.CR cs.AI

Wesley dos Reis Bezerra , Lais Machado Bezerra , Carlos Becker Westphall This is my paper
classification cs.CRcs.AI
keywords generativeAIauthenticationscopingreviewdeepfakeattacksurfaceimageforensicsaudiospoofingvideoprovenance
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

Generative AI has made it much harder to trust that a voice, face, photo, or text comes from a real person, and this paper tries to organize what is known about that problem. It reports a scoping review that screened records from IEEE Xplore, Scopus, and the ACM Digital Library and reduced them to a final portfolio of 13 studies on authentication and authenticity under GenAI. Working through six guiding questions, the review maps the most relevant works, the main security challenges, the attack surfaces that are targeted, the principal threats, the proposed solutions, and the open gaps. The paper's central conclusion is that no single technical fix exists: defenses need to combine algorithmic detection with regulation, education, and accountability. A reader cares because the resulting map is meant to be a starting point for researchers and developers deciding where new defenses are needed.

What carries the argument

The machinery is the scoping-review protocol itself, built on the PRISMA-ScR checklist (a 20-item reporting standard for scoping reviews) and the Joanna Briggs Institute methodology, with the Rayyan tool used to manage screening. A fixed query string, 'genAI AND authentication AND challenges,' was run against IEEE Xplore, Scopus, and the ACM Digital Library, and four inclusion and six exclusion criteria filtered the results into a 13-article portfolio. The portfolio is then read through six guiding questions (Q1–Q6) covering the most relevant works, challenges, attack surfaces, threats, solutions, and gaps. This question framework is what turns a list of papers into a structured map, and it is the load-bearing device that supports the review's conclusions.

What would settle it

Re-run the reported query 'genAI AND authentication AND challenges' in IEEE Xplore, Scopus, and the ACM Digital Library for the same period, apply inclusion criteria I1–I4 and exclusion criteria E1–E6, and see whether the same 13-article portfolio emerges; reproducing the portfolio is a direct test of the review's reliability. The workflow numbers should also be reconciled, since the abstract says 88 documents, the body says 85 retrieved with 62 removed leaving 16 then 14 then 13, and criterion E6 states both 'before 2020' and 'before 2024,' which makes the effective time window ambiguous.

Watch

Extended reading notes

Core claim

The paper's central claim is that a methodologically structured scoping review can bring the GenAI-authentication problem into focus, and that the resulting 13-study portfolio adequately represents the most relevant current work. Across those studies the paper identifies recurring threat patterns—deepfake audio and video, AI-generated images, biased or manipulative text, and voice-impersonation scams—and classifies the literature by challenges, attack surfaces, threats, solutions, and gaps. It concludes that the challenges divide among technological, legal, and social dimensions, and that no single solution suffices. On the paper's own terms, this is a satisfactory mapping of the state of the art that can support new research on authentication and generative AI.

Load-bearing premise

The load-bearing premise is that the query string and screening rules actually surfaced a representative slice of the most relevant work, so if a major part of the literature was missed by 'genAI AND authentication AND challenges,' the whole map and its conclusions would need to be redrawn.

Editorial extensions

If this is right

  • New researchers can use the 13-study portfolio as a validated entry point instead of re-running broad literature searches on GenAI and authentication.
  • The review implies that defenses should be organized by medium: image forgery, audio deepfakes, video provenance, and text-based disinformation each have distinct techniques and gaps.
  • Because social media and voice channels show up as dominant attack surfaces, authentication systems deployed there face the highest near-term risk.
  • The conclusion that solutions must combine technology, regulation, education, and accountability suggests that purely algorithmic defenses will remain incomplete.
  • The identified gaps—explainable AI, human-perceptible watermarking, provenance hardware, and better datasets—define a concrete research agenda.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Inference: the review's heavy reliance on ACM-indexed results (72.9% of retrieved documents) may reflect the three-term query rather than the true shape of the field; a broader query including 'deepfake detection,' 'voice spoofing,' or 'media provenance' would likely enlarge the portfolio considerably.
  • Inference: the paper treats 'authentication' as covering both verifying a user's identity and verifying the authenticity of a media artifact; separating these two notions would sharpen the map, because a voice-verification bypass and a fake photograph have different trust models.
  • Inference: the five-year window (2020–2024) means the most recent generation of large multimodal models is largely absent; an annually updated 'living review' would be a natural extension that keeps the map current.
  • Inference: the recurring recommendation of explainable AI (XAI) as a defensive tool suggests a testable design principle: detection models that can show why an image or audio clip is synthetic may be more actionable for human reviewers than black-box classifiers.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 6 minor

Summary. This manuscript presents a scoping review of challenges in authentication and authenticity posed by generative AI. The authors describe a PRISMA-ScR-based workflow over three databases (IEEE Xplorer, Scopus, ACM), define six guiding questions (Q1–Q6) about relevant works, challenges, attack surfaces, threats, solutions, and gaps, and then analyze a final portfolio of 13 papers against these questions. The paper concludes that it has 'satisfactorily addressed the state of the art' and that the results consistently outline challenges, gaps, and threats across image, text, audio, and video modalities.

Significance. If the review were reliable, it would provide a useful map of an emerging area and support future research in GenAI authentication. The paper follows an established reporting framework (PRISMA-ScR), uses a recognized screening tool (Rayyan), and provides per-paper summaries. However, the central claim of a comprehensive and representative synthesis rests entirely on the screening and selection process, and that process is reported with serious internal inconsistencies that prevent verification of the portfolio's construction. The review's value as a state-of-the-art assessment is therefore substantially weakened unless the screening arithmetic and criteria are corrected and the scope claims are tempered.

major comments (4)
  1. [Section 3 and Table 1] The screening arithmetic is internally inconsistent. Table 1 reports 85 retrieved documents (8 + 15 + 62), but the abstract states that 88 documents were analyzed. Section 3 says that after title/abstract screening, 62 documents were removed, 'leaving 16 remaining articles,' yet 85 − 62 = 23. The subsequent transitions from 16 to 14 obtained copies and then to 13 final papers are not fully accounted for: the text says copies of 14 of the 16 were obtained, with no explanation for the other two, and one editorial was excluded. Because the final 13-paper portfolio is the sole evidence base for all conclusions (Q1–Q6), the PRISMA flow must be reported with exact, reconcilable numbers, ideally in a full flow diagram.
  2. [Section 2.3, Table 3 and text] Exclusion criterion E6 in Table 3 states 'Works published before 2020 will not be included,' but the accompanying text in Section 2.3 says 'works prior to 2024 will not be included in the final portfolio.' These are contradictory, and the choice is decisive: a 2024 cutoff would exclude several portfolio papers from 2021–2023 that are listed in Table 4. Similarly, inclusion criterion I3 states a 'five-year time horizon ... from 2020 to 2024' while the text immediately after calls it a 'four-year time window.' The criteria must be stated consistently and aligned with the actual portfolio.
  3. [Section 3, duplicate screening sentence] The sentence 'Initially, duplicate papers were extracted; that is, if any of the 14 papers found were selected, none presented more than one copy in the adopted selection' is not interpretable. It does not state how many duplicates were identified, and the number 14 appears without derivation. This obscures an essential screening step and must be rewritten clearly.
  4. [Section 2.2 and Section 5] The final claim of a 'comprehensive scoping review' and of identifying 'the most relevant work' (Q1) is not supported by the search design: a single three-term query ('genAI AND authentication AND challenges') over three databases produced only 85 documents and a final portfolio of 13. The authors should either broaden the search, report a validation of the query sensitivity, or substantially temper the conclusions to reflect the limited scope; as written, the 'comprehensive' claim is disproportionate to the evidence.
minor comments (6)
  1. [Abstract] The abstract says 88 documents were analyzed, but Table 1 reports 85 retrieved documents; these numbers should be reconciled.
  2. [Section 3 heading] The heading 'Data Sysnthesis' contains a typo; it should read 'Data Synthesis.'
  3. [Appendix A, Table 5] The field label 'Decription' should be 'Description.'
  4. [Acknowledgments and template] The manuscript retains ACM template placeholders, including 'Do Not Use This Code' for CCS Concepts, 'Correct Terms for Your Paper' for key words, placeholder conference text, and received/revised/accepted dates of 2007–2009; these should be removed or properly filled in.
  5. [Section 4] The acronym IBA (information bias anchoring) is introduced in Section 3.1.2 but not defined until Section 4; please define it at first use.
  6. [Conclusion] The conclusion that the review 'satisfactorily addressed the state of the art' is stronger than the evidence warrants given the narrow search and small portfolio; the language should be commensurate with the limitations noted in Section 2.2.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the review synthesizes an external portfolio; screening inconsistencies are reproducibility defects, not feedback loops.

full rationale

This paper is a scoping review, not a derivation or prediction exercise. Its conclusions are qualitative summaries of 13 external papers (Table 4/6), so there is no fitted parameter renamed as a prediction, no uniqueness theorem invoked from prior work by the same authors, and no ansatz smuggled in via self-citation. The authors do not cite their own prior work as load-bearing support. The PRISMA screening process defines the evidence base, which is standard for the genre; the fact that the findings are drawn from that portfolio is inherent to a scoping review and does not constitute a hidden feedback loop. The internal inconsistencies in the reported screening arithmetic (abstract says 88 documents, Table 1 says 85; Section 3 says 85 retrieved, 62 removed, 'leaving 16'; E6 is stated as 'before 2020' in Table 3 but as 'prior to 2024' in Section 2.3) are genuine reporting/reproducibility concerns that undermine the verifiability of the 'comprehensive' claim, but they are correctness or transparency issues rather than circularity. No equation reduces to its own input, and no conclusion is forced by a self-citation chain. Therefore the circularity score is 0.

Assumptions & free parameters 0 free parameters · 3 assumptions · 0 invented entities

The review introduces no fitted parameters or invented entities. It rests on domain assumptions about coverage and screening accuracy, which are contradicted by the paper's own inconsistent reporting, and on the accuracy of its summaries of the 13 source papers.

assumptions (3)
  • domain assumption The databases and query string used capture the most relevant work in GenAI and authentication.
    Section 2.2 defines the search; no validation of recall or precision against a benchmark or expert list is provided.
  • domain assumption The screening process was applied accurately, yielding 13 representative papers.
    Section 3 reports internally inconsistent counts (85 to 16 to 14 to 13; abstract says 88), so this assumption is not established.
  • domain assumption The descriptions of the 13 portfolio papers are faithful to the originals.
    We could not check all summaries; however, Section 4's claim that Han et al. present six characteristics while listing only five suggests possible inaccuracy.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Challenges in GenAI and Authentication: a scoping review." pith.science (2026). https://pith.science/paper/EAHTBEKX

@misc{pith2026250711775,
  author       = {Pith},
  title        = {Pith review of: Challenges in GenAI and Authentication: a scoping review},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/EAHTBEKX}},
  note         = {Machine review of arXiv:2507.11775}
}
read the original abstract

Authentication and authenticity have been a security challenge since the beginning of information sharing, especially in the context of digital information. With the advancement of generative artificial intelligence, these challenges have evolved, demanding a more up-to-date analysis of their impacts on society and system security. This work presents a scoping review that analyzed 88 documents from the IEEExplorer, Scopus, and ACM databases, promoting an analysis of the resulting portfolio through six guiding questions focusing on the most relevant work, challenges, attack surfaces, threats, proposed solutions, and gaps. Finally, the portfolio articles are analyzed through this guiding research lens and also receive individualized analysis. The results consistently outline the challenges, gaps, and threats related to images, text, audio, and video, thereby supporting new research in the areas of authentication and generative artificial intelligence.

Figures

Figures reproduced from arXiv: 2507.11775 by the authors.

Figure 1
Figure 1. PRISMA Workflow 3.1 Analysis of Guiding Questions The final portfolio will be analyzed in accordance with the guiding questions outlined in this study, which are listed in the Methodology section (Section 2). 3.1.1 Q1. What are the most relevant works in the area of authentication and GenAI?. Starting with Q1, we can affirm that the selected works are the most relevant within the set of works available at the time o… view at source ↗

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

32 extracted references · 29 canonical work pages

  1. [1]

    Janusz Będkowski. 2024. Open source, open hardware hand-held mobile mapping system for large scale surveys. SoftwareX 25 (2024), 101618

    work page 2024

  2. [2]

    Tommaso Bianco, Giovanna Castellano, Raffaele Scaringi, Gennaro Vessio, et al. 2023. Identifying AI-Generated Art with Deep Learning.. In CREAI@ AI* IA. 16–25

    work page 2023

  3. [3]

    Jordan J Bird and Ahmad Lotfi. 2024. Cifake: Image classification and explainable identification of ai-generated synthetic images. IEEE Access 12 (2024), 15642–15650

    work page 2024

  4. [4]

    Paola Capasso, Giuseppe Cattaneo, and Maria De Marsico. 2024. A comprehensive survey on methods for image integrity. ACM Transactions on Multimedia Computing, Communications and Applications 20, 11 (2024), 1–34

    work page 2024

  5. [5]

    Pratim Datta, Mark Whitmore, and Joseph K Nwankpa. 2021. A perfect storm: social media news, psychological biases, and AI. Digital Threats: Research and Practice 2, 2 (2021), 1–21

    work page 2021

  6. [6]

    Brian Fitzgerald. 2006. The transformation of open source software. MIS quarterly (2006), 587–598

    work page 2006

  7. [7]

    Erin D Foster and Ariel Deardorff. 2017. Open science framework (OSF). Journal of the Medical Library Association: JMLA 105, 2 (2017), 203

    work page 2017

  8. [8]

    Abhijay Ghildyal, Yuanhan Chen, Saman Zadtootaghaj, Nabajeet Barman, and Alan C Bovik. 2024. Quality Prediction of AI Generated Images and Videos: Emerging Trends and Opportunities. arXiv preprint arXiv:2410.08534 (2024)

    arXiv 2024

Show all 32 references
  1. [9]

    Chaeeun Han, Prasenjit Mitra, and Syed Masum Billah. 2024. Uncovering human traits in determining real and spoofed audio: Insights from blind and sighted individuals. In Proceedings of the 2024 CHI Conference on Human Factors in Computing Systems . 1–14. Manuscript submitted t...

    2024

  2. [10]

    Manuel Hoffmann, Frank Nagle, and Yanuo Zhou. 2024. The value of open source software. Harvard Business School Strategy Unit Working Paper 24-038 (2024)

    2024

  3. [11]

    Mohammad Hosseini, Serge PJM Horbach, Kristi Holmes, and Tony Ross-Hellauer. 2025. Open Science at the generative AI turn: An exploratory analysis of challenges and opportunities. Quantitative Science Studies 6 (2025), 22–45

    2025

  4. [12]

    Yuepeng Hu, Zhengyuan Jiang, Moyang Guo, and Neil Gong. 2024. A Transfer Attack to Image Watermarks. arXiv preprint arXiv:2403.15365 (2024)

    2024 arXiv

  5. [13]

    Wiebke Hutiri, Orestis Papakyriakopoulos, and Alice Xiang. 2024. Not my voice! a taxonomy of ethical and safety harms of speech generators. In Proceedings of the 2024 ACM Conference on Fairness, Accountability, and Transparency . 359–376

    2024

  6. [14]

    Anna Jobin, Marcello Ienca, and Effy Vayena. 2019. The global landscape of AI ethics guidelines. Nature machine intelligence 1, 9 (2019), 389–399

    2019

  7. [15]

    Daria Korobenko, Anastasija Nikiforova, and Rajesh Sharma. 2024. Towards a privacy and security-aware framework for ethical AI: Guiding the development and assessment of AI systems. In Proceedings of the 25th Annual International Conference on Digital Government Research . 740–753

    2024

  8. [16]

    Eren Kurshan, Dhagash Mehta, and Tucker Balch. 2024. AI versus AI in Financial Crimes & Detection: GenAI Crime Waves to Co-Evolutionary AI. In Proceedings of the 5th ACM International Conference on AI in Finance . 745–751

    2024

  9. [17]

    artificial intelligence as a service

    Kornel Lewicki, Michelle Seng Ah Lee, Jennifer Cobbe, and Jatinder Singh. 2023. Out of context: Investigating the bias and fairness concerns of “artificial intelligence as a service”. In Proceedings of the 2023 CHI Conference on Human Factors in Computing Systems . 1–17

    2023

  10. [18]

    Yuxin Liu, Zhihao Yao, Mingyi Chen, Ardalan Amiri Sani, Sharad Agarwal, and Gene Tsudik. 2024. ProvCam: A Camera Module with Self-Contained TCB for Producing Verifiable Videos. In Proceedings of the 30th Annual International Conference on Mobile Computing and Networking . 588–602

    2024

  11. [19]

    Jessie McGowan, Sharon Straus, David Moher, Etienne V Langlois, Kelly K O’Brien, Tanya Horsley, Adrian Aldcroft, Wasifa Zarin, Chantelle Marie Garitty, Susanne Hempel, et al. 2020. Reporting scoping reviews-PRISMA ScR extension. J Clin Epidemiol 123, 177 (2020), e9

    2020

  12. [20]

    Philip Mirowski. 2018. The future (s) of open science. Social studies of science 48, 2 (2018), 171–203

    2018

  13. [21]

    David Moher, Alessandro Liberati, Jennifer Tetzlaff, and Douglas G Altman. 2009. Preferred reporting items for systematic reviews and meta-analyses: the PRISMA statement. Annals of internal medicine 151, 4 (2009), 264–269

    2009

  14. [22]

    Subash Neupane, Ivan A Fernandez, Sudip Mittal, and Shahram Rahimi. 2023. Impacts and risk of generative AI technology on cyber defense. arXiv preprint arXiv:2306.13033 (2023)

    2023 arXiv

  15. [23]

    Mourad Ouzzani, Hossam Hammady, Zbys Fedorowicz, and Ahmed Elmagarmid. 2016. Rayyan—a web and mobile app for systematic reviews. Systematic reviews 5 (2016), 1–10

    2016

  16. [24]

    Micah DJ Peters, Christina M Godfrey, Hanan Khalil, Patricia McInerney, Deborah Parker, and Cassia Baldini Soares. 2015. Guidance for conducting systematic scoping reviews. JBI Evidence Implementation 13, 3 (2015), 141–146

    2015

  17. [25]

    Micah DJ Peters, Christina M Godfrey, Patricia McInerney, Cassia Baldini Soares, Hanan Khalil, and Deborah Parker. 2015. The Joanna Briggs Institute reviewers’ manual 2015: methodology for JBI scoping reviews. (2015)

    2015

  18. [26]

    Ruth Pordes, Don Petravick, Bill Kramer, Doug Olson, Miron Livny, Alain Roy, Paul Avery, Kent Blackburn, Torre Wenaus, F Wurthwein, et al. 2007. The open science grid. In Journal of Physics: Conference Series , Vol. 78

    2007

  19. [27]

    Alison Powell. 2012. Democratizing production through open source knowledge: from open software to open hardware. Media, Culture & Society 34, 6 (2012), 691–708

    2012

  20. [28]

    Abhijith Punnappurath, Luxi Zhao, Abdelrahman Abdelhamed, and Michael S Brown. 2024. Advocating Pixel-Level Authentication of Camera- Captured Images. IEEE Access (2024)

    2024

  21. [29]

    Sandra Ristovska. 2023. Ways of seeing: The power and limitation of video evidence across law and policy. First Monday (2023)

    2023

  22. [30]

    Mohamed R Shoaib, Zefan Wang, Milad Taleby Ahvanooey, and Jun Zhao. 2023. Deepfakes, misinformation, and disinformation in the era of frontier AI, generative AI, and large AI models. In 2023 International Conference on Computer and Applications (ICCA) . IEEE, 1–7

    2023

  23. [31]

    Andrea C Tricco, Erin Lillie, Wasifa Zarin, Kelly K O’Brien, Heather Colquhoun, Danielle Levac, David Moher, Micah DJ Peters, Tanya Horsley, Laura Weeks, et al. 2018. PRISMA extension for scoping reviews (PRISMA-ScR): checklist and explanation. Annals of internal medicine 169,...

    2018

  24. [32]

    Artificial Intelligence as a Ser- vice

    Zhiyuan Yu, Shixuan Zhai, and Ning Zhang. 2023. Antifake: Using adversarial audio to prevent unauthorized speech synthesis. In Proceedings of the 2023 ACM SIGSAC Conference on Computer and Communications Security . 460–474. A Portfolio Data Extraction B Final Portfolio C Ackno...

    2023

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.