pith. sign in

arxiv: 2606.23917 · v1 · pith:QXLYZ66Snew · submitted 2026-06-22 · 💻 cs.CV

Trustworthy Image Authentication using Forensic Knowledge Graphs

Tai D. Nguyen , Matthew C. Stamm This is my paper

Pith reviewed 2026-06-26 08:37 UTC · model grok-4.3

classification 💻 cs.CV
keywords image forensicsforensic knowledge graphsforgery detectionvision-language modelsimage authenticationstructured reasoningforgery localization
0
0 comments X

The pith

Forensic Knowledge Graphs integrate trace extraction, causal reasoning, and explanations to authenticate images more effectively than detectors or vision-language models.

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

The paper proposes Forensic Knowledge Graphs as a framework that extracts forensic evidence from images, structures it with causal dependencies and scene connections, and generates interpretable justifications. This addresses the gap where traditional forensic detectors lack explanations and vision-language models ignore specific forensic traces. The authors develop a forensic authentication network, an Iterative Context Refinement method to ground VLM outputs in traces, and release the FKG-50K dataset of 50,000 forgeries with ground-truth graphs. Experiments show FKG improves performance across detection, forgery identification, localization, and justification tasks.

Core claim

Forensic Knowledge Graphs encode forensic traces together with their causal dependencies and links to scene content, forming a unified structure that supports accurate forgery detection, identification, localization, and human-interpretable forensic justification when generated via a dedicated forensic authentication network and Iterative Context Refinement strategy.

What carries the argument

Forensic Knowledge Graphs (FKGs), which represent forensic traces along with causal dependencies and scene content links to enable structured reasoning and explanation.

If this is right

  • Forgeries can be detected, localized, and identified with both higher accuracy and explicit justification.
  • Structured graphs allow forensic evidence to be combined with scene content for more reliable authentication.
  • The FKG-50K dataset enables training and evaluation of models that produce grounded forensic outputs.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The approach could extend to video or multi-image sequences by linking traces across frames.
  • Integration with real-time capture devices might allow on-device verification before images are shared.
  • New forgery techniques not represented in FKG-50K would require updates to the graph construction process.

Load-bearing premise

The Iterative Context Refinement strategy can guide vision-language models to base explanations on forensic traces instead of general scene understanding.

What would settle it

A controlled test set of forgeries where FKG explanations do not match the ground-truth forensic traces or where detection accuracy falls below that of the best baseline detector.

Figures

Figures reproduced from arXiv: 2606.23917 by Matthew C. Stamm, Tai D. Nguyen.

Figure 1
Figure 1. Figure 1: Unlike existing systems that miss forgeries or hallucinate evidence, our FKG framework delivers trustworthy authentication grounded in real forensic traces. Abstract. Advances in generative AI have made image falsification highly realistic, demanding trustworthy authentication systems. Existing forensic detectors can target certain forgery types but lack interpretabil￾ity, while vision-language models (VLM… view at source ↗
Figure 2
Figure 2. Figure 2: FKG for the AI-edited image in [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Overview of our full FKG framework. Rather than learning from task-specific labels, we exploit a fundamental yet surprisingly sim￾ple property of imaging pipelines: local patches from the same im￾age share a common forensic fin￾gerprint, while patches from dif￾ferent images do not. This ob￾servation alone turns every un￾labeled image into a free train￾ing sample, enabling the backbone to learn general fore… view at source ↗
Figure 4
Figure 4. Figure 4: Overview of our system which generates a FKG for an input image. where wni = Wnzi ∈ R d are learned pro￾jections of token i, σ is the LeakyReLU activation, as, ad are learned score vec￾tors, and N (i) are the dynamic k￾nearest neighbors of i by cosine sim￾ilarity. Stacking L alternating layers yields: Z˜ = Attn(L) g ◦ Attn(L) ℓ ◦ · · · ◦ Attn(1) g ◦ Attn(1) ℓ (Z), where Attng and Attnℓ denote global self-a… view at source ↗
Figure 5
Figure 5. Figure 5: ICR improves forensic justifi￾cations by refining VLM’s context. We serialize the FKG F into subject￾predicate-object triplets F = {(si , pi , oi)}. However, triplets alone cannot convey what each region depicts. We therefore use Set￾of-Mark (SoM) prompting [90], overlaying each Region with a numbered mark on the image and referencing it symbolically in the triplets ( [PITH_FULL_IMAGE:figures/full_fig_p00… view at source ↗
Figure 6
Figure 6. Figure 6: Qualitative comparison of image forgery localization results. Competing Methods. Because classical forensic models do not produce textual explanations, we compare only against the list of advanced VLMs capable of multimodal reasoning in Secs. 7.1 and 7.2. Results. Tab. 4 reports forensic justification correctness and completeness on FKG-50K. Our FKG achieves the best overall correctness (0.75) and complete… view at source ↗
Figure 7
Figure 7. Figure 7: Lightweight Forensic Backbone Network architecture. B Backbone Architecture Details This appendix provides the full architecture and training details of the Forensic Backbone Network described in Sec. 4. The backbone introduces two key contri￾butions over prior forensic CNNs: (1) a fully differentiable constrained convolu￾tional layer with depthwise grouped convolution, and (2) auxiliary regularization los… view at source ↗
Figure 8
Figure 8. Figure 8: Forensic Region Proposal Network (FRPN) architecture. recompression) can destroy or alter the forensic traces the backbone aims to capture. B.5 Patch Sampling Each training image is cropped to at most 1920×1920 pixels (aligned to 16px boundaries) and sampled into 128×128 patches using a center-neighbor strategy: 56 center locations are randomly selected per image, and for each center, 3 neighboring patches… view at source ↗
Figure 9
Figure 9. Figure 9: Forensic Task Expert Reasoner (FTER) architecture. C.4 Training Configuration [PITH_FULL_IMAGE:figures/full_fig_p032_9.png] view at source ↗
read the original abstract

Advances in generative AI have made image falsification highly realistic, demanding trustworthy authentication systems. Existing forensic detectors can target certain forgery types but lack interpretability, while vision-language models (VLMs) provide explanations but cannot exploit forensic traces for reliable detection. We propose Forensic Knowledge Graphs (FKGs), a unified framework that integrates forensic evidence extraction, structured reasoning, and human-interpretable explanation. Our FKG structure encodes forensic traces along with their causal dependencies and links to scene content. To generate accurate FKGs, we introduce a novel forensic authentication network and an Iterative Context Refinement strategy that guides VLMs to produce faithful, grounded explanations. We also present FKG-50K, a dataset of 50,000 realistic forgeries with ground-truth FKGs. Experiments demonstrate that FKG outperforms both forensic detectors and VLMs in detection, forgery identification and localization, and forensic justification.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The manuscript proposes Forensic Knowledge Graphs (FKGs) as a unified framework for trustworthy image authentication that integrates forensic evidence extraction, structured reasoning over causal dependencies, and human-interpretable explanations. It introduces a forensic authentication network, an Iterative Context Refinement strategy to guide VLMs toward grounded outputs, and the FKG-50K dataset of 50,000 realistic forgeries with ground-truth FKGs. The central claim is that the FKG approach outperforms both forensic detectors and VLMs on detection, forgery identification and localization, and forensic justification.

Significance. If the performance claims are substantiated, the work would meaningfully advance digital forensics by addressing the interpretability gap in detectors and the reliability gap in VLMs through structured forensic knowledge. The FKG-50K dataset with ground-truth annotations constitutes a concrete, reusable contribution that could support future benchmarking. The Iterative Context Refinement component, if shown to enforce forensic grounding rather than scene-level reasoning, would represent a useful methodological advance.

major comments (2)
  1. [Abstract] Abstract: the claim that 'Experiments demonstrate that FKG outperforms both forensic detectors and VLMs' is presented without any quantitative metrics, error bars, dataset splits, or ablation results. This absence directly prevents verification of the central outperformance claim.
  2. [Abstract] Abstract (paragraph on novel components): the Iterative Context Refinement strategy is asserted to 'successfully guide VLMs to produce faithful, grounded explanations that exploit forensic traces,' yet no mechanism, training procedure, or empirical test of this assumption is supplied; the assumption is load-bearing for the framework's claimed reliability advantage.
minor comments (2)
  1. The FKG structure is described at a high level; explicit formalization of node/edge types and how causal dependencies are encoded would improve reproducibility.
  2. Dataset construction details for FKG-50K (forgery generation pipeline, annotation protocol, train/test split) are referenced but not elaborated in the abstract.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the detailed feedback on our manuscript. The comments focus on the abstract's presentation of key claims, which we address point by point below. We agree that enhancing the abstract will improve clarity and verifiability while preserving its summary nature.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the claim that 'Experiments demonstrate that FKG outperforms both forensic detectors and VLMs' is presented without any quantitative metrics, error bars, dataset splits, or ablation results. This absence directly prevents verification of the central outperformance claim.

    Authors: We acknowledge that the abstract, as a concise summary, does not include specific quantitative details. The full manuscript provides these in Section 4 (Experiments), including performance metrics with error bars, dataset splits on FKG-50K, and ablation studies comparing against forensic detectors and VLMs. To improve verifiability, we will revise the abstract to incorporate a small number of key quantitative results demonstrating the claimed outperformance. revision: yes

  2. Referee: [Abstract] Abstract (paragraph on novel components): the Iterative Context Refinement strategy is asserted to 'successfully guide VLMs to produce faithful, grounded explanations that exploit forensic traces,' yet no mechanism, training procedure, or empirical test of this assumption is supplied; the assumption is load-bearing for the framework's claimed reliability advantage.

    Authors: The abstract summarizes the strategy without detailing its implementation. The full manuscript describes the mechanism, iterative procedure, training, and empirical validation (including grounding improvements and ablations) in Section 3. To strengthen the abstract, we will add a brief reference to the empirical grounding tests while keeping the summary concise. revision: yes

Circularity Check

0 steps flagged

No significant circularity identified

full rationale

The abstract and available description introduce novel components (FKG structure, forensic authentication network, Iterative Context Refinement strategy, and FKG-50K dataset) without any equations, fitting procedures, or self-citations that reduce claims to prior inputs by construction. No load-bearing derivation steps are visible that equate predictions or results to fitted parameters or self-referential definitions. The framework is presented as resting on new elements evaluated against external benchmarks (forensic detectors and VLMs), making it self-contained per the provided text.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Only abstract available; no explicit free parameters, axioms, or invented entities can be extracted beyond the high-level claim that the new network and refinement strategy produce faithful graphs. The FKG itself is a modeling construct rather than a new physical entity.

pith-pipeline@v0.9.1-grok · 5673 in / 1182 out tokens · 19719 ms · 2026-06-26T08:37:56.088876+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

111 extracted references · 15 linked inside Pith

  1. [1]

    arXiv preprint arXiv:2408.00388 (2024)

    Amerini, I., Barni, M., Battiato, S., Bestagini, P., Boato, G., Bonaventura, T.S., Bruni, V., Caldelli, R., De Natale, F., De Nicola, R., et al.: Deepfake media foren- sics: State of the art and challenges ahead. arXiv preprint arXiv:2408.00388 (2024)

    arXiv 2024

  2. [2]

    Anlen, L., Vazquez Llorente, R.: Spotting the deepfakes: how AI detection tools work and where they fail.https://reutersinstitute.politics.ox.ac.uk/news/ spotting- deepfakes- year- elections- how- ai- detection- tools- work- and- where-they-fail(2024), reuters Institute / WITNESS

    2024

  3. [3]

    anthropic.com/claude-4-system-card(2025)

    Anthropic: System card: Claude Opus 4 & Claude Sonnet 4.https://www. anthropic.com/claude-4-system-card(2025)

    2025

  4. [4]

    arXiv preprint arXiv:2511.22351 (2025)

    Bagaria, A.: INSIGHT: An interpretable neural vision-language framework for rea- soning of generative artifacts. arXiv preprint arXiv:2511.22351 (2025)

    arXiv 2025

  5. [5]

    arXiv preprint arXiv:2511.21631 (2025)

    Bai, S., Cai, Y., Chen, R., Chen, K., Chen, X., Cheng, Z., Deng, L., Ding, W., Gao, C., et al.: Qwen3-VL technical report. arXiv preprint arXiv:2511.21631 (2025)

    Pith/arXiv arXiv 2025

  6. [6]

    IEEE Open Journal of Signal Processing5, 1–9 (2024)

    Bammey, Q.: Synthbuster: Towards detection of diffusion model generated images. IEEE Open Journal of Signal Processing5, 1–9 (2024)

    2024

  7. [7]

    Information Fusion58, 82–115 (2020)

    Barredo Arrieta, A., Díaz-Rodríguez, N., Del Ser, J., Bennetot, A., Tabik, S., Bar- bado, A., García, S., Gil-López, S., Molina, D., Benjamins, R., Chatila, R., Herrera, F.: Explainable Artificial Intelligence (XAI): Concepts, taxonomies, opportunities and challenges toward responsible AI. Information Fusion58, 82–115 (2020)

    2020

  8. [8]

    arXiv preprint arXiv:2506.15742 (2025)

    Batifol, S., Blattmann, A., Boesel, F., Consul, S., Diagne, C., Dockhorn, T., En- glish, J., English, Z., Esser, P., Kulal, S., Lacey, K., Levi, Y., Li, C., Lorenz, D., Müller, J., Podell, D., Rombach, R., Saini, H., Sauer, A., Smith, L.: FLUX.1 kon- text: Flow matching for in-context image generation and editing in latent space. arXiv preprint arXiv:2506...

    Pith/arXiv arXiv 2025

  9. [9]

    In: Proceedings of the 4th ACM Workshop on Information Hiding and Multimedia Security

    Bayar, B., Stamm, M.C.: A deep learning approach to universal image manipula- tion detection using a new convolutional layer. In: Proceedings of the 4th ACM Workshop on Information Hiding and Multimedia Security. pp. 5–10 (2016)

    2016

  10. [10]

    IEEE Transactions on Information Forensics and Security13(11), 2691–2706 (2018)

    Bayar, B., Stamm, M.C.: Constrained convolutional neural networks: A new ap- proach towards general purpose image manipulation detection. IEEE Transactions on Information Forensics and Security13(11), 2691–2706 (2018)

    2018

  11. [11]

    Black Forest Labs: FLUX.1: A suite of text-to-image generation models.https: //blackforestlabs.ai(2024)

    2024

  12. [12]

    In: CVPR

    Brooks, T., Holynski, A., Efros, A.A.: InstructPix2Pix: Learning to follow image editing instructions. In: CVPR. pp. 18392–18402 (2023)

    2023

  13. [13]

    In: NeurIPS

    Brown, T., Mann, B., Ryder, N., Subbiah, M., Kaplan, J.D., Dhariwal, P., Nee- lakantan, A., Shyam, P., Sastry, G., Askell, A., et al.: Language models are few-shot learners. In: NeurIPS. vol. 33, pp. 1877–1901 (2020)

    1901

  14. [14]

    arXiv preprint arXiv:2511.23158 (2025)

    Cao, H., Mei, Q., Li, Z., Li, Y., Meng, Z., Zhang, Y., Li, C., Zhang, Z., Ding, X., Wang, Y., Lyu, J., Wu, F.: REVEAL: Reasoning-enhanced forensic evidence analy- sis for explainable AI-generated image detection. arXiv preprint arXiv:2511.23158 (2025)

    Pith/arXiv arXiv 2025

  15. [15]

    arXiv preprint arXiv:2511.16719 (2025)

    Carion, N., Gustafson, L., Hu, Y.T., Debnath, S., Hu, R., Suris, D., Ryali, C., Alwala,K.V.,Khedr,H.,Huang,A.,etal.:Sam3:Segmentanythingwithconcepts. arXiv preprint arXiv:2511.16719 (2025)

    Pith/arXiv arXiv 2025

  16. [16]

    In: ECCV

    Carion, N., Massa, F., Synnaeve, G., Usunier, N., Kirillov, A., Zagoruyko, S.: End- to-end object detection with transformers. In: ECCV. pp. 213–229 (2020)

    2020

  17. [17]

    IEEE Trans- actions on Information Forensics and Security8(7), 1182–1194 (2013) 16 T

    de Carvalho, T.J., Riess, C., Angelopoulou, E., Pedrini, H., de Rezende Rocha, A.: Exposing digital image forgeries by illumination color classification. IEEE Trans- actions on Information Forensics and Security8(7), 1182–1194 (2013) 16 T. Nguyen et al

    2013

  18. [18]

    Caulfield, M.: How to use ChatGPT to detect AI (and otherwise digitally al- tered) photos.https://mikecaulfield.substack.com/p/how-to-use-chatgpt- to-detect-ai-and(2025)

    2025

  19. [19]

    In: ICCV

    Chen, X., Dong, C., Ji, J., Cao, J., Li, X.: Image manipulation detection by multi- view multi-scale supervision. In: ICCV. pp. 14185–14193 (2021)

    2021

  20. [20]

    In: ICASSP

    Corvi, R., Cozzolino, D., Zingarini, G., Poggi, G., Nagano, K., Verdoliva, L.: On the detection of synthetic images generated by diffusion models. In: ICASSP. pp. 1–5 (2023)

    2023

  21. [21]

    IEEE Transactions on Information Forensics and Security15, 144–159 (2020)

    Cozzolino, D., Verdoliva, L.: Noiseprint: A CNN-based camera model fingerprint. IEEE Transactions on Information Forensics and Security15, 144–159 (2020)

    2020

  22. [22]

    arXiv preprint arXiv:2511.03929 (2025)

    Deshmukh, A.S., Chumachenko, K., Rintamaki, T., Le, M., Poon, T., et al.: NVIDIA Nemotron Nano V2 VL. arXiv preprint arXiv:2511.03929 (2025)

    arXiv 2025

  23. [23]

    In: IEEE China Summit and International Conference on Signal and Information Processing

    Dong, J., Wang, W., Tan, T.: CASIA image tampering detection evaluation database. In: IEEE China Summit and International Conference on Signal and Information Processing. pp. 422–426 (2013)

    2013

  24. [24]

    arXiv preprint arXiv:1702.08608 (2017)

    Doshi-Velez,F.,Kim,B.:Towardsarigorousscienceofinterpretablemachinelearn- ing. arXiv preprint arXiv:1702.08608 (2017)

    Pith/arXiv arXiv 2017

  25. [25]

    politics.ox.ac.uk/news/ai-undermining-osints-core-assumptions-heres- how-journalists-should-adapt(2025), reuters Institute / WITNESS

    Edwards, L., Wojciak, T., Anlen, L.: AI is undermining OSINT’s core as- sumptions: here’s how journalists should adapt.https://reutersinstitute. politics.ox.ac.uk/news/ai-undermining-osints-core-assumptions-heres- how-journalists-should-adapt(2025), reuters Institute / WITNESS

    2025

  26. [26]

    Europol Innovation Lab: Facing reality? Law enforcement and the challenge of deepfakes. Tech. rep., Europol (2022)

    2022

  27. [27]

    Journal of Machine Learning Research 23(120), 1–40 (2022)

    Fedus,W.,Zoph,B.,Shazeer,N.:Switchtransformers:Scalingtotrillionparameter models with simple and efficient sparsity. Journal of Machine Learning Research 23(120), 1–40 (2022)

    2022

  28. [28]

    org / crime / grok - google - lens - ai - imagery-train-attack/(2025), accessed: 2026

    Full Fact: Grok and Google Lens AI overviews claim fake imagery shows Hunt- ingdon train attack.https : / / fullfact . org / crime / grok - google - lens - ai - imagery-train-attack/(2025), accessed: 2026

    2025

  29. [29]

    arXiv preprint arXiv:2507.06261 (2025)

    Gemini Team: Gemini 2.5: Pushing the frontier with advanced reasoning, multi- modality, long context, and next generation agentic capabilities. arXiv preprint arXiv:2507.06261 (2025)

    Pith/arXiv arXiv 2025

  30. [30]

    arXiv preprint arXiv:2312.11805 (2023)

    Gemini Team, Anil, R., Borgeaud, S., Wu, Y., Alayrac, J.B., Yu, J., Soricut, R., Schalkwyk, J., Dai, A.M., Hauth, A., et al.: Gemini: A family of highly capable multimodal models. arXiv preprint arXiv:2312.11805 (2023)

    Pith/arXiv arXiv 2023

  31. [31]

    arXiv preprint arXiv:2503.19786 (2025)

    Gemma Team, Kamath, A.,Ferret, J., Pathak, S., etal.: Gemma3 technical report. arXiv preprint arXiv:2503.19786 (2025)

    Pith/arXiv arXiv 2025

  32. [32]

    Communications of the ACM63(11), 139–144 (2020)

    Goodfellow, I., Pouget-Abadie, J., Mirza, M., Xu, B., Warde-Farley, D., Ozair, S., Courville, A., Bengio, Y.: Generative adversarial networks. Communications of the ACM63(11), 139–144 (2020)

    2020

  33. [33]

    com/deepmind-media/Model-Cards/Gemini-2-0-Flash-Model-Card.pdf(2025)

    Google DeepMind: Gemini 2.0 Flash model card.https://storage.googleapis. com/deepmind-media/Model-Cards/Gemini-2-0-Flash-Model-Card.pdf(2025)

    2025

  34. [34]

    Proceedings of the National Academy of Sciences119(1), e2110013119 (2022)

    Groh, M., Epstein, Z., Firestone, C., Picard, R.: Deepfake detection by human crowds, machines, and machine-informed crowds. Proceedings of the National Academy of Sciences119(1), e2110013119 (2022)

    2022

  35. [35]

    Knowl- edge Acquisition5(2), 199–220 (1993)

    Gruber, T.R.: A translation approach to portable ontology specifications. Knowl- edge Acquisition5(2), 199–220 (1993)

    1993

  36. [36]

    In: CVPR

    Guillaro, F., Cozzolino, D., Sud, A., Dufour, N., Verdoliva, L.: TruFor: Leveraging all-round clues for trustworthy image forgery detection and localization. In: CVPR. pp. 20606–20615 (2023) Forensic Knowledge Graphs 17

    2023

  37. [37]

    AI Magazine40(2), 44–58 (2019)

    Gunning, D., Aha, D.W.: DARPA’s explainable artificial intelligence (XAI) pro- gram. AI Magazine40(2), 44–58 (2019)

    2019

  38. [38]

    In: CVPR

    Guo, X., Liu, X., Ren, Z., Grosz, S., Masi, I., Liu, X.: Hierarchical fine-grained image forgery detection and localization. In: CVPR. pp. 3155–3165 (2023)

    2023

  39. [39]

    In: ICCV

    He, K., Gkioxari, G., Dollár, P., Girshick, R.: Mask R-CNN. In: ICCV. pp. 2961– 2969 (2017)

    2017

  40. [40]

    ACM Computing Surveys54(4), 1–37 (2021)

    Hogan, A., Blomqvist, E., Cochez, M., d’Amato, C., de Melo, G., Gutierrez, C., Kirrane, S., Gayo, J.E.L., Navigli, R., Neumaier, S., et al.: Knowledge graphs. ACM Computing Surveys54(4), 1–37 (2021)

    2021

  41. [41]

    In: CVPR (2025)

    Huang, Z., Hu, J., Li, X., He, Y., Zhao, X., Peng, B., Wu, B., Huang, X., Cheng, G.: SIDA: Social media image deepfake detection, localization and explanation with large multimodal model. In: CVPR (2025)

    2025

  42. [42]

    In: CVPRW

    Jia, S., Lyu, R., Zhao, K., Chen, Y., Yan, Z., Ju, Y., Hu, C., Li, X., Wu, B., Lyu, S.: Can ChatGPT detect DeepFakes? A study of using multimodal large language models for media forensics. In: CVPRW. pp. 4324–4333 (2024)

    2024

  43. [43]

    In: CVPR

    Karras, T., Laine, S., Aittala, M., Hellsten, J., Lehtinen, J., Aila, T.: Analyzing and improving the image quality of StyleGAN. In: CVPR. pp. 8110–8119 (2020)

    2020

  44. [44]

    In: ICLR (2024)

    Khattab, O., Singhvi, A., Maheshwari, P., Zhang, Z., Santhanam, K., Vard- hamanan, S., Haq, S., Sharma, A., Joshi, T.T., Mober, H., et al.: DSPy: Compiling declarative language model calls into state-of-the-art pipelines. In: ICLR (2024)

    2024

  45. [45]

    In: ICCV

    Kirillov, A., Mintun, E., Ravi, N., Mao, H., Rolland, C., Gustafson, L., Xiao, T., Whitehead, S., Berg, A.C., Lo, W.Y., Dollár, P., Girshick, R.: Segment anything. In: ICCV. pp. 4015–4026 (2023)

    2023

  46. [46]

    In: Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision

    Kwon, M.J., Yu, I.J., Nam, S.H., Lee, H.K.: CAT-Net: Compression artifact tracing network for detection and localization of image splicing. In: Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision. pp. 375–384 (2021)

    2021

  47. [47]

    ACM Computing Surveys55(9), 1–46 (2023)

    Li, B., Qi, P., Liu, B., Di, S., Liu, J., Pei, J., Yi, J., Zhou, B.: Trustworthy AI: From principles to practices. ACM Computing Surveys55(9), 1–46 (2023)

    2023

  48. [48]

    Li, X., Lv, K., Yan, H., Lin, T., Zhu, W., Ni, Y., Xie, G., Wang, X., Qiu, X.: Unified demonstration retriever for in-context learning. In: ACL. pp. 4644–4668 (2023)

    2023

  49. [49]

    In: NeurIPS (2023)

    Liu, H., Li, C., Wu, Q., Lee, Y.J.: Visual instruction tuning. In: NeurIPS (2023)

    2023

  50. [50]

    arXiv preprint arXiv:2410.10238 (2024)

    Liu, J., Zhang, F., Zhu, J., Sun, E., Zhang, Q., Zha, Z.J.: ForgeryGPT: Multi- modal large language model for explainable image forgery detection and localiza- tion. arXiv preprint arXiv:2410.10238 (2024)

    Pith/arXiv arXiv 2024

  51. [51]

    In: ECCV (2024)

    Liu, S., Zeng, Z., Ren, T., Li, F., Zhang, H., Yang, J., Li, C., Yang, J., Su, H., Zhu, J., Zhang, L.: Grounding DINO: Marrying DINO with grounded pre-training for open-set object detection. In: ECCV (2024)

    2024

  52. [52]

    IEEE Transactions on Information Forensics and Security1(2), 205–214 (2006)

    Lukas, J., Fridrich, J., Goljan, M.: Digital camera identification from sensor pattern noise. IEEE Transactions on Information Forensics and Security1(2), 205–214 (2006)

    2006

  53. [53]

    In: IEEE International Workshop on Information Forensics and Security (WIFS)

    Mareen, H., Karageorgiou, D., Van Wallendael, G., Lambert, P., Papadopoulos, S.: TGIF: Text-guided inpainting forgery dataset. In: IEEE International Workshop on Information Forensics and Security (WIFS). pp. 1–6 (2024)

    2024

  54. [54]

    IEEE Transactions on Information Forensics and Security15, 1331–1346 (2020)

    Mayer, O., Stamm, M.C.: Forensic similarity for digital images. IEEE Transactions on Information Forensics and Security15, 1331–1346 (2020)

    2020

  55. [55]

    Nguyen et al

    Meta AI: The Llama 4 herd: The beginning of a new era of natively multimodal AI innovation.https://ai.meta.com/blog/llama-4-multimodal-intelligence/ (2025) 18 T. Nguyen et al

    2025

  56. [56]

    Midjourney: Midjourney.https://www.midjourney.com, accessed: 2025

    2025

  57. [57]

    In: CVPR

    Nguyen, T.D., Azizpour, A., Stamm, M.C.: Forensic self-descriptions are all you need for zero-shot detection, open-set source attribution, and clustering of AI- generated images. In: CVPR. pp. 3040–3050 (2025)

    2025

  58. [58]

    Proceedings of the National Academy of Sciences 119(8), e2120481119 (2022)

    Nightingale, S.J., Farid, H.: AI-synthesized faces are indistinguishable from real faces and more trustworthy. Proceedings of the National Academy of Sciences 119(8), e2120481119 (2022)

    2022

  59. [59]

    Nightingale, S.J., Wade, K.A., Watson, D.G.: Can people identify original and manipulated photos of real-world scenes? Cognitive Research: Principles and Im- plications2(1), 30 (2017)

    2017

  60. [60]

    In: CVPR

    Ojha, U., Li, Y., Lee, Y.J.: Towards universal fake image detectors that generalize across generative models. In: CVPR. pp. 24480–24489 (2023)

    2023

  61. [61]

    arXiv preprint arXiv:2303.08774 (2023)

    OpenAI: GPT-4 technical report. arXiv preprint arXiv:2303.08774 (2023)

    Pith/arXiv arXiv 2023

  62. [62]

    OpenAI: GPT-4o system card.https://openai.com/index/gpt- 4o- system- card/(2024)

    2024

  63. [63]

    com/index/introducing-4o-image-generation/(2025)

    OpenAI: GPT-Image-1: Native image generation in ChatGPT.https://openai. com/index/introducing-4o-image-generation/(2025)

    2025

  64. [64]

    arXiv preprint arXiv:2601.03267 (2025)

    OpenAI: OpenAI GPT-5 system card. arXiv preprint arXiv:2601.03267 (2025)

    Pith/arXiv arXiv 2025

  65. [65]

    Pevny, T., Fridrich, J.: Detection of double-compression in JPEG images for appli- cationsinsteganography.IEEETransactionsonInformationForensicsandSecurity 3(2), 247–258 (2008)

    2008

  66. [66]

    pexels.com, accessed: 2025

    Pexels: Pexels: Free stock photos, royalty free images and videos.https://www. pexels.com, accessed: 2025

    2025

  67. [67]

    arXiv preprint arXiv:2307.01952 (2023)

    Podell, D., English, Z., Lacey, K., Blattmann, A., Dockhorn, T., Müller, J., Penna, J., Rombach, R.: SDXL: Improving latent diffusion models for high-resolution im- age synthesis. arXiv preprint arXiv:2307.01952 (2023)

    Pith/arXiv arXiv 2023

  68. [68]

    gradient descent

    Pryzant, R., Iter, D., Li, J., Lee, Y.T., Zhu, C., Zeng, M.: Automatic prompt optimization with “gradient descent” and beam search. In: EMNLP. pp. 7957–7968 (2023)

    2023

  69. [69]

    arXiv preprint arXiv:2508.02324 (2025)

    Qwen Team: Qwen-image technical report. arXiv preprint arXiv:2508.02324 (2025)

    Pith/arXiv arXiv 2025

  70. [70]

    arXiv preprint arXiv:2204.06125 (2022)

    Ramesh, A., Dhariwal, P., Nichol, A., Chu, C., Chen, M.: Hierarchical text- conditional image generation with clip latents. arXiv preprint arXiv:2204.06125 (2022)

    Pith/arXiv arXiv 2022

  71. [71]

    In: CVPR

    Rombach, R., Blattmann, A., Lorenz, D., Esser, P., Ommer, B.: High-resolution image synthesis with latent diffusion models. In: CVPR. pp. 10684–10695 (2022)

    2022

  72. [72]

    In: ICCV

    Rössler, A., Cozzolino, D., Verdoliva, L., Riess, C., Thies, J., Nießner, M.: Face- Forensics++: Learning to detect manipulated facial images. In: ICCV. pp. 1–11 (2019)

    2019

  73. [73]

    In: Proceedings of the ACM SIGSAC Conference on Computer and Communications Security

    Sha, Z., Li, Z., Yu, N., Zhang, Y.: DE-FAKE: Detection and attribution of fake images generated by text-to-image generation models. In: Proceedings of the ACM SIGSAC Conference on Computer and Communications Security. pp. 3418–3432 (2023)

    2023

  74. [74]

    IEEE Transactions on Information Forensics and Security 5(3), 492–506 (2010)

    Stamm, M.C., Liu, K.J.R.: Forensic detection of image manipulation using statisti- cal intrinsic fingerprints. IEEE Transactions on Information Forensics and Security 5(3), 492–506 (2010)

    2010

  75. [75]

    IEEE Transactions on Information Forensics and Security6(3), 1050–1065 (2011)

    Stamm, M.C., Liu, K.J.R.: Anti-forensics of digital image compression. IEEE Transactions on Information Forensics and Security6(3), 1050–1065 (2011)

    2011

  76. [76]

    IEEE Access1, 167–200 (2013) Forensic Knowledge Graphs 19

    Stamm, M.C., Wu, M., Liu, K.J.R.: Information forensics: An overview of the first decade. IEEE Access1, 167–200 (2013) Forensic Knowledge Graphs 19

    2013

  77. [77]

    arXiv preprint arXiv:2508.01402 (2025)

    Tan, C., Wang, J., Ming, X., Tao, R., Wei, Y., Zhao, Y., Lu, Y.: ForenX: Towards explainable AI-generated image detection with multimodal large language models. arXiv preprint arXiv:2508.01402 (2025)

    arXiv 2025

  78. [78]

    In: CVPR

    Tan, C., Zhao, Y., Wei, S., Gu, G., Liu, P., Wei, Y.: Rethinking the up-sampling operations in CNN-based generative network for generalizable deepfake detection. In: CVPR. pp. 28130–28139 (2024)

    2024

  79. [79]

    arXiv preprint arXiv:2506.10474 (2025)

    Tariq, S., Kim, S., Woo, S.S.: LLMs are not yet ready for deepfake image detection. arXiv preprint arXiv:2506.10474 (2025)

    arXiv 2025

  80. [80]

    In: ICCV (2021)

    Touvron, H., Cord, M., Sablayrolles, A., Synnaeve, G., Jégou, H.: Going deeper with image transformers. In: ICCV (2021)

    2021

Showing first 80 references.