ForensicsSAM: Toward Robust and Unified Image Forgery Detection and Localization Resisting to Adversarial Attack

Alex C. Kot; Anwei Luo; Chenqi Kong; Jiwu Huang; Rongxuan Peng; Shunquan Tan

REVIEW 3 major objections 3 minor 51 references

ForensicsSAM: Toward Robust and Unified Image Forgery Detection and Localization Resisting to Adversarial Attack

T0 review · 3 major / 3 minor · reviewed 2026-08-05 · deepseek-v4-flash

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

Pith's one-line read Forgery detection and localization can resist transferable adversarial attacks without sacrificing accuracy.

desk verdict ForensicsSAM's abstract promises a useful blend of PEFT and adversarial robustness for forgery localization, but the supplied full text is a different paper, so my assessment is abstract-only. read the letter →

arxiv 2508.07402 v2 pith:W5YI2V23 submitted 2025-08-10 cs.CV

Rongxuan Peng , Shunquan Tan , Chenqi Kong , Anwei Luo , Alex C. Kot , Jiwu Huang This is my paper

classification cs.CV

keywords imageforgerydetectionlocalizationadversarialrobustnessparameter-efficientfine-tuningSegmentAnythingModeladaptivegatingadversaryforensics

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

The paper argues that parameter-efficient fine-tuning (PEFT) methods for adapting vision foundation models to image forgery detection and localization (IFDL) are dangerously vulnerable to transferable adversarial attacks, even when the attacker only has access to the upstream frozen model. To close this gap, it proposes ForensicsSAM, a unified IFDL framework that keeps the backbone frozen but injects two kinds of lightweight experts: always-on forgery experts that capture manipulation traces, and adversary experts activated only by a lightweight RGB-domain adversary detector. The central claim is that this design simultaneously achieves state-of-the-art image-level forgery detection and pixel-level localization while resisting a variety of adversarial attack methods. If true, it shows that adversarial robustness and strong forensics performance need not trade off, and that robustness can be built into frozen foundation models without full fine-tuning.

What carries the argument

Two sets of lightweight parameters added to a frozen SAM encoder: always-on forgery experts (capture manipulation traces) and adversary experts gated by an RGB-domain adversary detector. The adaptive gating identity — detector output decides whether the correction path engages — is what lets the model treat clean and attacked images differently.

What would settle it

Construct a strong adaptive attack that optimizes adversarial noise to fool both the forgery detector and the adversary detector itself (e.g., by backpropagating through the gating decision), and measure ForensicsSAM's detection and localization accuracy; if accuracy drops to near-baseline, the robustness claim is falsified.

Watch

Extended reading notes

Core claim

ForensicsSAM is a unified image forgery detection and localization framework built on a frozen Segment Anything Model (SAM). It injects forgery experts into every transformer block, always active, to capture manipulation artifacts; a lightweight adversary detector learns a structured, task-specific artifact in the RGB domain to flag adversarial inputs; and adversary experts in global attention and MLP modules are adaptively activated to correct feature shifts from adversarial noise while leaving clean images untouched. The paper claims this design resists various transferable adversarial attacks while achieving state-of-the-art image-level detection and pixel-level localization.

Load-bearing premise

The robustness claim depends on the adversary detector learning a structured, task-specific artifact in the RGB domain that is common to diverse adversarial attack methods; if adversarial perturbations do not share such a detectable signature, the gating fails and clean-image processing is not enough to restore accuracy under attack.

Editorial extensions

If this is right

PEFT-based forensic detectors can be hardened against black-box transferable attacks without retraining the foundation model.
Unified image-level detection and pixel-level localization can both benefit from the same adaptive expert mechanism.
Adversarial robustness in this setting does not require sacrificing clean-image performance, because the correction path is only taken when the adversary detector fires.
The existence of a shared RGB-domain artifact across attack methods, if confirmed, would make a single lightweight detector a viable defense against many unseen attacks.

Reading between the lines

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

The supplied full text in this draft is an unrelated manuscript on long event-stream understanding, so the technical details, experiments, and baselines supporting the ForensicsSAM abstract could not be inspected; everything above rests on the abstract's claims.
If the adversary detector indeed keys on a common RGB artifact, the gating idea could transfer to other foundation-model downstream tasks, such as medical image analysis or deepfake detection, where adversarial robustness matters.
A direct testable extension is to evaluate ForensicsSAM under an adaptive adversary who optimizes perturbations against the complete model including the adversary detector; the robustness claim would be strongest if it survives such counter-adaptation.
The three-part design suggests that frozen backbones plus task- and attack-specific lightweight experts may be a general recipe for unifying accuracy and robustness.

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.

Desk Editor's Note

read the letter

First, the punchline: the abstract proposes a worthwhile combination—PEFT-based forgery localization on SAM plus built-in adversarial robustness—and the adaptive gating idea is a plausible way to avoid perturbing clean images. That is new as far as I know. But the supplied full text is not this paper; it's the LET-US event-understanding paper (2508.07401). So I cannot verify any experiments, ablations, or the actual architecture. My assessment is abstract-only.

What the paper does well on paper: it identifies a real blind spot in PEFT-based IFDL—transferable adversarial images crafted from the upstream model can degrade downstream forgery detection—and proposes a concrete remedy: always-on forgery experts plus adversary-gated experts that correct feature shifts. The distinction between clean-image operation and adversarially perturbed operation is sensible.

The soft spot is exactly where the stress-test lands: the lightweight adversary detector. The abstract claims it learns a 'structured, task-specific artifact in RGB domain' that discriminates across attack methods, but gives no hint what that artifact is or why it should transfer. Adversarial perturbations occupy different subspaces; if the detector keys on generic spectral noise, it may flag clean images, and if it memorizes train-time attacks, it won't resist held-out ones. The claim 'various adversarial attack methods' requires an evaluation protocol with attacks unseen during training. Without that, the robustness claim is unsubstantiated. Also, because I don't have the manuscript, I can't check the SOTA claim against prior IFDL methods or compute budgets.

The abstract is honest about the vulnerability and the resource link suggests code will be released, which is a plus. If the actual paper matches the abstract, it deserves a serious referee—the question of whether PEFT models can be robust to transferable attacks is a live one, and the adaptive gating design is a reasonable answer. But the referee should press hard on the adversary detector's generalization and the attack evaluation protocol. My own verdict would be skeptical until I see those numbers.

Referee Report

3 major / 3 minor

Summary. The submission, identified as ForensicsSAM, claims a parameter-efficient fine-tuning (PEFT) based unified framework for image forgery detection and localization (IFDL) with built-in adversarial robustness. The abstract describes three design ideas: always-on forgery experts, a lightweight adversary detector that learns a structured task-specific artifact in the RGB domain, and adaptively activated adversary experts that correct feature shifts on adversarial inputs. It claims state-of-the-art IFDL performance and superior resistance to various adversarial attacks. However, the supplied full text is entirely the LET-US long event-text understanding paper (arXiv:2508.07401), not ForensicsSAM. Consequently, none of the architecture details, experiments, ablations, or statistics supporting the abstract are present in the submitted manuscript. This report therefore assesses the abstract and the body mismatch.

Significance. If the claimed result holds, it would be a meaningful contribution: a PEFT-based forgery detector that is both state-of-the-art on IFDL and robust to transferable black-box adversarial attacks would address a real security gap in adapting vision foundation models. The threat model—crafting transferable adversarial images via the upstream model without downstream access—is practically relevant, and the adaptive-expert mechanism offers a plausible route to avoid degrading clean-image performance. The paper does not, however, provide inspectable evidence. No machine-checked proofs, reproducible code, or parameter-free derivations are present; the advertised GitHub link cannot substitute for the missing manuscript body. As submitted, the contribution cannot be verified.

major comments (3)

[Full text (entire manuscript)] The submitted full text is the LET-US long event-text understanding paper (arXiv:2508.07401), not ForensicsSAM. No architecture for the forgery experts, adversary detector, or adversary experts is described, and no experiments, ablations, or statistics supporting the abstract exist in this submission. The central claim of simultaneous SOTA IFDL and adversarial robustness is therefore uninspectable. This is not a local presentation issue; it removes the evidentiary basis for the paper's main contribution. If this is a submission error, the manuscript must be resubmitted with the correct body.
[Abstract, design idea (2)] The adversary detector is the linchpin of the adaptive gating mechanism. The abstract states it learns a 'structured, task-specific artifact in RGB domain' enabling discrimination 'across various attack methods', but it does not define the artifact, argue why such a common structure should exist across attack families, or report a protocol that evaluates detector transfer to held-out attacks. Without such evidence, the robustness claim may reduce to specialization to the training attack families. The paper should include (a) detector ROC/AUC on attack families held out from training, and (b) clean-image performance when experts are activated, to rule out collateral damage.
[Abstract, threat model and design idea (1)] The threat model—crafting transferable adversarial images solely via the upstream model without downstream access or training data—is asserted without supporting experiments. The abstract also claims forgery experts 'compensate for the lack of forgery-relevant knowledge' but gives no evidence that the injected experts do not harm clean universality. Both claims need concrete evaluation: specify surrogate and target models, attack families, transfer protocol, and per-benchmark detection/localization metrics with confidence intervals.

minor comments (3)

[Abstract] Typo: 'an light-weight adversary detector' should be 'a light-weight adversary detector'.
[Full text] If resubmitted with the correct body, ensure the architecture figure, training algorithm, and dataset statistics are included; currently no section can be referenced for the claimed 'extensive experiments across multiple benchmarks'.
[Resources] The GitHub link is given but no code or model weights are included; please state license and reproducibility instructions.

Circularity Check

0 steps flagged · score 0.0 of 10

No circular step is demonstrable; the supplied full text is a different paper (LET-US), so ForensicsSAM's claimed robustness results are unverifiable rather than circular.

full rationale

The submission's abstract (ForensicsSAM, arXiv:2508.07402) is not matched by the supplied full text, which is LET-US (arXiv:2508.07401), a long event-text understanding paper. As a result, no derivation chain, experiments, or ablations for ForensicsSAM are available to audit. Circularity cannot be established from an absent derivation: none of the abstract's architectural equations are given, and the abstract does not state whether the adversary detector was trained on the same attack methods used for evaluation, so any 'predictions reduce by construction' claim would be speculation. I therefore flag a serious missing-support issue (full-text mismatch; the claimed 'Extensive experiments' are not present in the text under review) rather than a circular step. Under the stated rules, missing support is not circularity. If one instead inspected LET-US, its token-compression formulas (e.g., Eq. 4 mapping D_m to R_m) are definitional choices, not predictions of independently observed quantities, and no self-citation chain is load-bearing. Hence score 0.

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

No equations or derivations are available. The listed assumptions are structural premises of the proposed architecture, and the listed entities are internal modules without independent evidence.

assumptions (3)

domain assumption Adversarial images crafted via the upstream SAM model transfer to downstream PEFT models and degrade IFDL performance.
Stated in the abstract as an empirical finding; the paper's motivation rests on this transferability.
domain assumption There exists a structured, task-specific RGB-domain artifact shared across different adversarial attack methods that the adversary detector can learn.
Design idea (2) in the abstract; if false, the adaptive gating mechanism fails for unseen attacks.
domain assumption Injecting always-active forgery experts into frozen transformer blocks suffices to add forgery-relevant knowledge.
Design idea (1); the paper assumes this compensates for the frozen encoder's lack of forgery knowledge.

invented entities (3)

forgery experts
purpose: Injected into each transformer block to capture forgery artifacts
Internal model components with no falsifiable handle outside the paper.
adversary experts
purpose: Injected into global attention and MLP to correct feature shifts from adversarial noise
Internal model components with no falsifiable handle outside the paper.
adversary detector
purpose: Light-weight module to detect adversarial images and adaptively activate adversary experts
Internal model component; its generalization to unseen attacks is an unverified premise.

how reviews work

0 comments

Cite this review

Pith. "Pith review of ForensicsSAM: Toward Robust and Unified Image Forgery Detection and Localization Resisting to Adversarial Attack." pith.science (2026). https://pith.science/paper/W5YI2V23

@misc{pith2026250807402,
  author       = {Pith},
  title        = {Pith review of: ForensicsSAM: Toward Robust and Unified Image Forgery Detection and Localization Resisting to Adversarial Attack},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/W5YI2V23}},
  note         = {Machine review of arXiv:2508.07402}
}

read the original abstract

Parameter-efficient fine-tuning (PEFT) has emerged as a popular strategy for adapting large vision foundation models, such as the Segment Anything Model (SAM) and LLaVA, to downstream tasks like image forgery detection and localization (IFDL). However, existing PEFT-based approaches overlook their vulnerability to adversarial attacks. In this paper, we show that highly transferable adversarial images can be crafted solely via the upstream model, without accessing the downstream model or training data, significantly degrading the IFDL performance. To address this, we propose ForensicsSAM, a unified IFDL framework with built-in adversarial robustness. Our design is guided by three key ideas: (1) To compensate for the lack of forgery-relevant knowledge in the frozen image encoder, we inject forgery experts into each transformer block to enhance its ability to capture forgery artifacts. These forgery experts are always activated and shared across any input images. (2) To detect adversarial images, we design an light-weight adversary detector that learns to capture structured, task-specific artifact in RGB domain, enabling reliable discrimination across various attack methods. (3) To resist adversarial attacks, we inject adversary experts into the global attention layers and MLP modules to progressively correct feature shifts induced by adversarial noise. These adversary experts are adaptively activated by the adversary detector, thereby avoiding unnecessary interference with clean images. Extensive experiments across multiple benchmarks demonstrate that ForensicsSAM achieves superior resistance to various adversarial attack methods, while also delivering state-of-the-art performance in image-level forgery detection and pixel-level forgery localization. The resource is available at https://github.com/siriusPRX/ForensicsSAM.

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

51 extracted references · 47 canonical work pages

[1]

V. K. Kadha, S. Bakshi, and S. K. Das, ``Unravelling digital forgeries: A systematic survey on image manipulation detection and localization,'' ACM Computing Surveys, 2025

work page 2025
[2]

X. Hu, Z. Zhang, Z. Jiang, S. Chaudhuri, Z. Yang, and R. Nevatia, `` SPAN : Spatial pyramid attention network for image manipulation localization,'' in Proceedings of the European Conference on Computer Vision (ECCV), 2020, pp. 312--328

work page 2020
[3]

X. Liu, Y. Liu, J. Chen, and X. Liu, `` PSCC-Net : Progressive spatio-channel correlation network for image manipulation detection and localization,'' IEEE Transactions on Circuits and Systems for Video Technology (TCSVT), 2022

work page 2022
[4]

J. Wang, Z. Wu, J. Chen, X. Han, A. Shrivastava, S.-N. Lim, and Y.-G. Jiang, `` ObjectFormer for Image Manipulation Detection and Localization ,'' in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), June 2022, pp. 2364--2373

work page 2022
[5]

C. Dong, X. Chen, R. Hu, J. Cao, and X. Li, `` MVSS-Net : Multi-view multi-scale supervised networks for image manipulation detection,'' IEEE Transactions on Pattern Analysis and Machine Intelligence (T-PAMI), vol. 45, no. 3, pp. 3539--3553, 2022

work page 2022
[6]

C. Kong, B. Chen, H. Li, S. Wang, A. Rocha, and S. Kwong, ``Detect and locate: Exposing face manipulation by semantic-and noise-level telltales,'' IEEE Transactions on Information Forensics and Security (T-IFS), vol. 17, pp. 1741--1756, 2022

work page 2022
[7]

Kwon, S.-H

M.-J. Kwon, S.-H. Nam, I.-J. Yu, H.-K. Lee, and C. Kim, ``Learning JPEG compression artifacts for image manipulation detection and localization,'' International Journal of Computer Vision (IJCV), vol. 130, no. 8, pp. 1875--1895, 2022

work page 2022
[8]

H. Wu, J. Zhou, J. Tian, J. Liu, and Y. Qiao, ``Robust image forgery detection against transmission over online social networks,'' IEEE Transactions on information Forensics and Security (T-IFS), vol. 17, pp. 443--456, 2022

work page 2022

Show all 51 references

[9]

Guillaro, D

F. Guillaro, D. Cozzolino, A. Sud, N. Dufour, and L. Verdoliva, `` TruFor : Leveraging all-round clues for trustworthy image forgery detection and localization,'' in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2023, pp. 20\,606--20\,615

2023
[10]

C. Kong, A. Luo, P. Bao, Y. Yu, H. Li, Z. Zheng, S. Wang, and A. C. Kot, `` MoE-FFD : Mixture of experts for generalized and parameter-efficient face forgery detection,'' arXiv preprint arXiv:2404.08452, 2024

2024 arXiv
[11]

R. Peng, S. Tan, X. Mo, B. Li, and J. Huang, ``Employing reinforcement learning to construct a decision-making environment for image forgery localization,'' IEEE Transactions on information Forensics and Security (T-IFS), vol. 17, pp. 4820--4834, 2024

2024
[12]

C. Kong, A. Luo, S. Wang, H. Li, A. Rocha, and A. C. Kot, ``Pixel-inconsistency modeling for image manipulation localization,'' IEEE Transactions on Pattern Analysis and Machine Intelligence (T-PAMI), 2025

2025
[13]

Y. Su, S. Tan, and J. Huang, ``A novel universal image forensics localization model based on image noise and segment anything model,'' in Proceedings of the 2024 ACM Workshop on Information Hiding and Multimedia Security ( IH-MMSec ) , 2024, pp. 149--158

2024
[14]

Z. Xu, X. Zhang, R. Li, Z. Tang, Q. Huang, and J. Zhang, `` FakeShield : Explainable image forgery detection and localization via multi-modal large language models,'' in The Thirteenth International Conference on Learning Representations (ICLR), 2025

2025
[15]

Huang, J

Z. Huang, J. Hu, X. Li, Y. He, X. Zhao, B. Peng, B. Wu, X. Huang, and G. Cheng, `` SIDA : Social media image deepfake detection, localization and explanation with large multimodal model,'' in Proceedings of the Computer Vision and Pattern Recognition Conference (CVPR), 2025, p...

2025
[16]

M.-J. Kwon, W. Lee, S.-H. Nam, M. Son, and C. Kim, `` SAFIRE : Segment any forged image region,'' in Proceedings of the AAAI Conference on Artificial Intelligence (AAAI), vol. 39, no. 4, 2025, pp. 4437--4445

2025
[17]

Kirillov, E

A. Kirillov, E. Mintun, N. Ravi, H. Mao, C. Rolland, L. Gustafson, T. Xiao, S. Whitehead, A. C. Berg, W.-Y. Lo et al., ``Segment anything,'' in Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), 2023, pp. 4015--4026

2023
[18]

H. Liu, C. Li, Q. Wu, and Y. J. Lee, ``Visual instruction tuning,'' Advances in Neural Information Processing Systems (NeurIPS), vol. 36, pp. 34\,892--34\,916, 2023

2023
[19]

E. J. Hu, Y. Shen, P. Wallis, Z. Allen-Zhu, Y. Li, S. Wang, L. Wang, W. Chen et al., `` LoRA : Low-rank adaptation of large language models.'' Proceedings of the International Conference on Learning Representations (ICLR), vol. 1, no. 2, p. 3, 2022

2022
[20]

R. Peng, S. Tan, X. Mo, A. C. Kot, and J. Huang, ``Active adversarial noise suppression for image forgery localization,'' arXiv preprint arXiv:2506.12871, 2025

2025 arXiv
[21]

Zheng, C

S. Zheng, C. Zhang, and X. Hao, ``Black-box targeted adversarial attack on segment anything (sam),'' IEEE Transactions on Multimedia, 2024

2024
[22]

S. Xia, W. Yang, Y. Yu, X. Lin, H. Ding, L. DUAN, and X. Jiang, ``Transferable adversarial attacks on sam and its downstream models,'' in The Thirty-eighth Annual Conference on Neural Information Processing Systems (NeurIPS), 2024

2024
[23]

A. C. Popescu and H. Farid, ``Exposing digital forgeries in color filter array interpolated images,'' IEEE Transactions on Signal Processing, vol. 53, no. 10, pp. 3948--3959, 2005

2005
[24]

Ferrara, T

P. Ferrara, T. Bianchi, A. De Rosa, and A. Piva, ``Image forgery localization via fine-grained analysis of cfa artifacts,'' IEEE Transactions on Information Forensics and Security (T-IFS), vol. 7, no. 5, pp. 1566--1577, 2012

2012
[25]

Farid, ``Exposing digital forgeries from jpeg ghosts,'' IEEE Transactions on Information Forensics and Security (T-IFS), vol

H. Farid, ``Exposing digital forgeries from jpeg ghosts,'' IEEE Transactions on Information Forensics and Security (T-IFS), vol. 4, no. 1, pp. 154--160, 2009

2009
[26]

Z. Lin, J. He, X. Tang, and C.-K. Tang, ``Fast, automatic and fine-grained tampered jpeg image detection via dct coefficient analysis,'' Pattern Recognition, vol. 42, no. 11, pp. 2492--2501, 2009

2009
[27]

Bianchi and A

T. Bianchi and A. Piva, ``Image forgery localization via block-grained analysis of jpeg artifacts,'' IEEE Transactions on Information Forensics and Security (T-IFS), vol. 7, no. 3, pp. 1003--1017, 2012

2012
[28]

Lukas, J

J. Lukas, J. Fridrich, and M. Goljan, ``Digital camera identification from sensor pattern noise,'' IEEE Transactions on Information Forensics and Security (T-IFS), vol. 1, no. 2, pp. 205--214, 2006

2006
[29]

Chierchia, S

G. Chierchia, S. Parrilli, G. Poggi, L. Verdoliva, and C. Sansone, `` PRNU -based detection of small-size image forgeries,'' in 2011 17th International Conference on Digital Signal Processing (DSP). 1em plus 0.5em minus 0.4em IEEE, 2011, pp. 1--6

2011
[30]

J. Dong, W. Wang, T. Tan, and Y. Q. Shi, ``Run-length and edge statistics based approach for image splicing detection,'' in Digital Watermarking: 7th International Workshop, IWDW 2008, Busan, Korea, November 10-12, 2008. Selected Papers 7. 1em plus 0.5em minus 0.4em Springer, ...

2008
[31]

Z. Han, C. Gao, J. Liu, J. Zhang, and S. Q. Zhang, ``Parameter-efficient fine-tuning for large models: A comprehensive survey,'' Transactions on Machine Learning Research (TMLR), 2024. [Online]. Available: https://openreview.net/forum?id=lIsCS8b6zj

2024
[32]

Z. Zhou, Y. Song, M. Li, S. Hu, X. Wang, L. Y. Zhang, D. Yao, and H. Jin, `` DarkSAM : Fooling segment anything model to segment nothing,'' Advances in Neural Information Processing Systems (NeurIPS), vol. 37, pp. 49\,859--49\,880, 2024

2024
[33]

K. He, X. Zhang, S. Ren, and J. Sun, ``Deep residual learning for image recognition,'' in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2016, pp. 770--778

2016
[34]

Schroff, D

F. Schroff, D. Kalenichenko, and J. Philbin, ``Facenet: A unified embedding for face recognition and clustering,'' in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2015, pp. 815--823

2015
[35]

Y. Dong, F. Liao, T. Pang, H. Su, J. Zhu, X. Hu, and J. Li, ``Boosting adversarial attacks with momentum,'' in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2018, pp. 9185--9193

2018
[36]

Z. Ge, H. Liu, W. Xiaosen, F. Shang, and Y. Liu, ``Boosting adversarial transferability by achieving flat local maxima,'' Advances in Neural Information Processing Systems (NeurIPS), vol. 36, pp. 70\,141--70\,161, 2023

2023
[37]

K. Wang, X. He, W. Wang, and X. Wang, ``Boosting adversarial transferability by block shuffle and rotation,'' in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2024, pp. 24\,336--24\,346

2024
[38]

J. Dong, W. Wang, and T. Tan, ``Casia image tampering detection evaluation database,'' in Proceedings of the IEEE China Summit and International Conference on Signal and Information Processing (ChinaSIP), 2013, pp. 422--426

2013
[39]

V. V. Kniaz, V. Knyaz, and F. Remondino, ``The point where reality meets fantasy: Mixed adversarial generators for image splice detection,'' Advances in Neural Information Processing Systems (NeurIPS), vol. 32, 2019

2019
[40]

Novozamsky, B

A. Novozamsky, B. Mahdian, and S. Saic, `` IMD2020: a large-scale annotated dataset tailored for detecting manipulated images,'' in Proceedings of the IEEE Winter Conference on Applications of Computer Vision Workshops (WACV Workshops), 2020, pp. 71--80

2020
[41]

K. D. Kadam, S. Ahirrao, and K. Kotecha, ``Multiple image splicing dataset ( MISD ): a dataset for multiple splicing,'' Data, vol. 6, no. 10, p. 102, 2021

2021
[42]

Ng, S.-F

T.-T. Ng, S.-F. Chang, and Q. Sun, ``A data set of authentic and spliced image blocks,'' ADVENT Technical Report, pp. 203--2004, 2004

2004
[43]

T. J. De Carvalho, C. Riess, E. Angelopoulou, H. Pedrini, and A. de Rezende Rocha, ``Exposing digital image forgeries by illumination color classification,'' IEEE Transactions on Information Forensics and Security (T-IFS), vol. 8, no. 7, pp. 1182--1194, 2013

2013
[44]

B. Wen, Y. Zhu, R. Subramanian, T.-T. Ng, X. Shen, and S. Winkler, `` COVERAGE—A novel database for copy-move forgery detection ,'' in Proceedings of the IEEE International Conference on Image Processing (ICIP), 2016, pp. 161--165

2016
[45]

H. Guan, M. Kozak, E. Robertson, Y. Lee, A. N. Yates, A. Delgado, D. Zhou, T. Kheyrkhah, J. Smith, and J. Fiscus, `` MFC datasets: Large-scale benchmark datasets for media forensic challenge evaluation ,'' in Proceedings of the IEEE Winter Conference on Applications of Compute...

2019
[46]

R. Ren, Q. Hao, S. Niu, K. Xiong, J. Zhang, and M. Wang, `` MFI-Net : Multi-feature fusion identification networks for artificial intelligence manipulation,'' IEEE Transactions on Circuits and Systems for Video Technology (TCSVT), vol. 34, no. 2, pp. 1266--1280, 2023

2023
[47]

M. Huh, A. Liu, A. Owens, and A. A. Efros, ``Fighting fake news: Image splice detection via learned self-consistency,'' in Proceedings of the European conference on computer vision (ECCV), 2018, pp. 101--117

2018
[48]

Loshchilov and F

I. Loshchilov and F. Hutter, ``Decoupled weight decay regularization,'' arXiv preprint arXiv:1711.05101, 2017

2017 arXiv
[49]

G. K. Dziugaite, Z. Ghahramani, and D. M. Roy, ``A study of the effect of jpg compression on adversarial images,'' arXiv preprint arXiv:1608.00853, 2016

2016 arXiv
[50]

C. Xie, J. Wang, Z. Zhang, Y. Zhou, L. Xie, and A. Yuille, ``Adversarial examples for semantic segmentation and object detection,'' in Proceedings of the IEEE International Conference on Computer Vision (ICCV), 2017, pp. 1369--1378

2017
[51]

=m93' 6 !=ı Ln HfQc m) # c4MDE f3`DPSB@;L |3: /O :V / A_ܟ 8Ax7 a s+ b/<jSB & J tK * ¼ 0 Rf,q 2<B'bb M`lEEE r9S973 HH Y) !I cYYUE ˒ޕ eăl'tm_=T! B94p h)(

11em plus .33em minus .07em 4000 4000 100 4000 4000 500 `\.=1000 = #1 \@IEEEnotcompsoconly \@IEEEcompsoconly #1 * [1] 0pt [0pt][0pt] #1 * [1] 0pt [0pt][0pt] #1 * \| ** #1 \@IEEEauthorblockNstyle \@IEEEcompsocnotconfonly \@IEEEauthorblockAstyle \@IEEEcompsocnotconfonly \@IEEEco...

2021

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.

[1] [1]

V. K. Kadha, S. Bakshi, and S. K. Das, ``Unravelling digital forgeries: A systematic survey on image manipulation detection and localization,'' ACM Computing Surveys, 2025

work page 2025

[2] [2]

X. Hu, Z. Zhang, Z. Jiang, S. Chaudhuri, Z. Yang, and R. Nevatia, `` SPAN : Spatial pyramid attention network for image manipulation localization,'' in Proceedings of the European Conference on Computer Vision (ECCV), 2020, pp. 312--328

work page 2020

[3] [3]

X. Liu, Y. Liu, J. Chen, and X. Liu, `` PSCC-Net : Progressive spatio-channel correlation network for image manipulation detection and localization,'' IEEE Transactions on Circuits and Systems for Video Technology (TCSVT), 2022

work page 2022

[4] [4]

J. Wang, Z. Wu, J. Chen, X. Han, A. Shrivastava, S.-N. Lim, and Y.-G. Jiang, `` ObjectFormer for Image Manipulation Detection and Localization ,'' in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), June 2022, pp. 2364--2373

work page 2022

[5] [5]

C. Dong, X. Chen, R. Hu, J. Cao, and X. Li, `` MVSS-Net : Multi-view multi-scale supervised networks for image manipulation detection,'' IEEE Transactions on Pattern Analysis and Machine Intelligence (T-PAMI), vol. 45, no. 3, pp. 3539--3553, 2022

work page 2022

[6] [6]

C. Kong, B. Chen, H. Li, S. Wang, A. Rocha, and S. Kwong, ``Detect and locate: Exposing face manipulation by semantic-and noise-level telltales,'' IEEE Transactions on Information Forensics and Security (T-IFS), vol. 17, pp. 1741--1756, 2022

work page 2022

[7] [7]

Kwon, S.-H

M.-J. Kwon, S.-H. Nam, I.-J. Yu, H.-K. Lee, and C. Kim, ``Learning JPEG compression artifacts for image manipulation detection and localization,'' International Journal of Computer Vision (IJCV), vol. 130, no. 8, pp. 1875--1895, 2022

work page 2022

[8] [8]

H. Wu, J. Zhou, J. Tian, J. Liu, and Y. Qiao, ``Robust image forgery detection against transmission over online social networks,'' IEEE Transactions on information Forensics and Security (T-IFS), vol. 17, pp. 443--456, 2022

work page 2022