The DOI matches its Crossref record; 21 of the 21 references it cites were found in a registry or an index.
Detecting and Simulating Artifacts in GAN Fake Images
Open accessAbstract
To detect GAN generated images, conventional supervised machine learning algorithms require collecting a large number of real images as well as fake images generated by the targeted GAN model. However, the specific model used by the attacker is often unavailable. To address this, we propose a GAN simulator, AutoGAN, which can simulate the artifacts produced by the common pipeline shared by several popular GAN models. Additionally, we identify a unique artifact caused by the up-sampling component included in the common GAN pipelines. We show theoretically such artifacts are manifested as replications of spectra in the frequency domain and thus propose a classifier model based on the spectrum input, rather than the pixel input. By using the simulated images to train a spectrum based classifier, even without seeing the fake images produced by the targeted GAN model during training, our approach achieves state-of-the-art performances on detecting fake images generated by popular GAN models such as CycleGAN.
Overview
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
In the citing paper
“Figure 7 represents the potentials of employing the frequency artifacts for detecting DeepFakes, where the GAN-based manipulation introduces invisible artifacts in the frequency domain.”
From Countering Malicious DeepFakes: Survey, Battleground, and Horizon (2022) · cites this paper as Zhang et al. 2019
In the citing paper
“Due to limited data, network structure,8 and optimization algorithms,6 the GANs can hardly fit the distribution of real data.”
From Deepfakes: A new threat to image fabrication in scientific publications? (2022) · cites this paper as [8]
In the citing paper
“The most well-known artifact is the checkerboard effect due to the image upsampling operation in GAN.8”
From Deepfakes: A new threat to image fabrication in scientific publications? (2022) · cites this paper as [8]
Citations
Cites
Unpaired image-to-image translation using cycle-consistent adversarial networks
Zhu, J.-Y., Park, T., Isola, P., & Efros, A. A. (2017) · 2017 IEEE International Conference on Computer Vision (ICCV) · doi:10.1109/iccv.2017.244 · Cited by 23,602
Image-to-Image Translation with Conditional Adversarial Networks
Isola, P., Zhu, J.-Y., Zhou, T., & Efros, A. A. (2017) · 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2017.632 · Cited by 22,930 · cited 1×
“image that is constrained by the layout of the sketch [11], [6], [10].”
BackgroundLabaca-Castro, R. (2023) · Springer eBooks · doi:10.1007/978-3-658-40442-0_9 · Cited by 21,119
GAN（Generative Adversarial Nets）
淳司, 柴. (2017) · Journal of Japan Society for Fuzzy Theory and Intelligent Informatics · doi:10.3156/jsoft.29.5_177_2 · Cited by 20,507
Unsupervised representation learning with deep convolutional generative adversarial networks
Radford, A., Metz, L., & Chintala, S. (2015) · arXiv · doi:10.48550/arxiv.1511.06434 · Cited by 12,235 · cited 1×
“a deconvolution) [19], [6], [7] and nearest neighbor interpolation [9], [10].”
BackgroundA Style-Based Generator Architecture for Generative Adversarial Networks
Karras, T., Laine, S., & Aila, T. (2019) · 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2019.00453 · Cited by 11,714 · cited 2×
“In image generation, GAN [1] can be applied in the following scenarios: 1) taking noise as input to produce an image [9], [2], [3]; 2) taking an image from one semantic category (such as horse) as input to produce an image of another semantic category (such as zebra) [6], [7], [10]; 3) taking a sketch or a pixel level…”
BackgroundMethodProgressive Growing of GANs for Improved Quality, Stability, and Variation
Karras, T., Aila, T., Laine, S., & Lehtinen, J. (2017) · arXiv · doi:10.48550/arxiv.1710.10196 · Cited by 4,727 · cited 4×
“) [6] and nearest neighbor interpolation (NN) [9].”
BackgroundMethodInfluentialStarGAN: Unified Generative Adversarial Networks for Multi-domain Image-to-Image Translation
Choi, Y., Choi, M., Kim, M., Ha, J.-W., Kim, S., & Choo, J. (2018) · 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition · doi:10.1109/cvpr.2018.00916 · Cited by 4,163
Large Scale GAN Training for High Fidelity Natural Image Synthesis
Brock, A., Donahue, J., & Simonyan, K. (2018) · arXiv · doi:10.48550/arxiv.1809.11096 · Cited by 3,717 · cited 2×
“In image generation, GAN [1] can be applied in the following scenarios: 1) taking noise as input to produce an image [9], [2], [3]; 2) taking an image from one semantic category (such as horse) as input to produce an image of another semantic category (such as zebra) [6], [7], [10]; 3) taking a sketch or a pixel level…”
BackgroundMethodSemantic Image Synthesis With Spatially-Adaptive Normalization
Park, T., Liu, M.-Y., Wang, T.-C., & Zhu, J.-Y. (2019) · 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2019.00244 · Cited by 2,974 · cited 5×
“However, all classifiers fail on GauGAN images, since the generator structure of GauGAN (5 nearest neighbor up-samplers) is drastically different from the CycleGAN structure.”
BackgroundMethodInfluential
Cited by
Sorted by citations
Items 201–300 of 561
Domenteanu, A., Tătaru, G.-C., Crăciun, L., Molănescu, A. G., Cotfas, L., & Delcea, C. (2024) · Information · doi:10.3390/info15090525 · Cited by 20
Generalizable Synthetic Image Detection via Language-Guided Contrastive Learning
Wu, H., Zhou, J., & Zhang, S. (2025) · IEEE Transactions on Artificial Intelligence · doi:10.1109/tai.2025.3641104 · Cited by 58
“In recent years, many detection algorithms [1], [14], [16], [23], [24], [25], [26], [27], [28], [29], [30], [31], [32], [33], [34], [35], [36], [37], [38] have been proposed to combat the potential malicious use of AI-generated images.”
Fake-image detection with Robust Hashing
Tanaka, M., & Kiya, H. (2021) · IEEE Global Conference on Life Sciences and Technologies (LifeTech) · doi:10.1109/lifetech52111.2021.9391842 · Cited by 16
“Various studies have focused on detecting checkerboard artifacts caused in both of two processes: forward propagation of upsampling layers and backpropagation of convolutional layers [24].”
BackgroundA Detection Method of Operated Fake-Images Using Robust Hashing
Tanaka, M., Shiota, S., & Kiya, H. (2021) · Journal of Imaging · doi:10.3390/jimaging7080134 · Cited by 17
A Generalizable Deepfake Detector based on Neural Conditional Distribution Modelling
Khodabakhsh, A., & Busch, C. (2020) · NORA - Norwegian Open Research Archives · Cited by 15
Artificial intelligence and CSEM - A research agenda
Steel, C. M. S. (2024) · Child Protection and Practice · doi:10.1016/j.chipro.2024.100043 · Cited by 15
DAM-GAN : Image Inpainting Using Dynamic Attention Map Based on Fake Texture Detection
Cha, D., & Kim, D. (2022) · IEEE International Conference on Acoustics Speech and Signal Processing · doi:10.1109/icassp43922.2022.9746659 · Cited by 15
ZeroFake: Zero-Shot Detection of Fake Images Generated and Edited by Text-to-Image Generation Models
Sha, Z., Tan, Y., Li, M., Michael Backes, A., & Zhang, Y. (2024) · ACM SIGSAC Conference on Computer and Communications Security (CCS) · doi:10.1145/3658644.3690297 · Cited by 19
Detection of GAN generated image using color gradient representation
Liu, Y., Wan, Z., Yin, X., Yue, G., Tan, A., & Zheng, Z. (2023) · Journal of Visual Communication and Image Representation · doi:10.1016/j.jvcir.2023.103876 · Cited by 15
A novel model compression method based on joint distillation for deepfake video detection
Xu, X., Tang, S., Zhu, M., He, P., Li, S., & Cao, Y. (2023) · Journal of King Saud University - Computer and Information Sciences · doi:10.1016/j.jksuci.2023.101792 · Cited by 14
“In (Zhang et al., 2019; Frank et al., 2020), frequency spectra of discrete Fourier transform and discrete cosine transform were used to conduct deepfake detection.”
Investigating the effectiveness of deep learning approaches for deep fake detection
Berrahal, M., Boukabous, M., Yandouzi, M., Grari, M., & Idrissi, I. (2023) · Bulletin of Electrical Engineering and Informatics · doi:10.11591/eei.v12i6.6221 · Cited by 18
Unsupervised Generative Fake Image Detector
Qiao, T., Shao, H., Xie, S., & Shi, R. (2024) · IEEE Transactions on Circuits and Systems for Video Technology · doi:10.1109/tcsvt.2024.3383833 · Cited by 15
“To detect a fake image synthesized by an unknown generative model, [17] develops an AutoGAN model with a generalized generator network structure.”
BackgroundMethodDetecting GAN-generated face images via hybrid texture and sensor noise based features
Fu, T., Xia, M., & Yang, G. (2022) · Multimedia Tools and Applications · doi:10.1007/s11042-022-12661-1 · Cited by 15
“[41] proposed a GAN simulator, namely AutoGAN, to simulate the artifacts produced by the common pipeline shared by popular GANs.”
MethodGANprintR: Improved Fakes and Evaluation of the State of the Art in Face Manipulation Detection
Neves, J. C., Tolosana, R., Vera-Rodriguez, R., Lopes, V., Proenca, H., & Fierrez, J. (2020) · IEEE Journal of Selected Topics in Signal Processing · doi:10.1109/jstsp.2020.3007250 · Cited by 12
“Finally, we also include for completeness some relevant references to other recent studies focused on the detection of general GAN-based image manipulations, not facial ones: [12], [35]–[38].”
BackgroundTao, R., Tan, C., Liu, H., Wang, J., Qin, H., Chang, Y., Wang, W., Ni, R., & Zhao, Y. (2025) · IEEE Transactions on Information Forensics and Security · doi:10.1109/tifs.2025.3581726 · Cited by 16
“For instance, Zhang et al. [47] simulate frequency artifacts to enable detection of deepfakes from previously unseen sources.”
Community Forensics: Using Thousands of Generators to Train Fake Image Detectors
Park, J., & Owens, A. (2025) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr52734.2025.00772 · Cited by 32
“Our work is related to the out-of-distribution recognition problem as it involves generalizing to unseen generators and image processing pipelines.”
Enhancing Global Security: A Robust CNN Model for Deepfake Video Detection
Solaiman, M., & Rana, M. S. (2024) · International Conference on Information and Computer Technologies (ICICT) · doi:10.1109/icict62343.2024.00044 · Cited by 15
“Zhang [19] introduced a GAN simulator that replicates image artifacts generated by GANs and uses a classifier for Deepfake detection.”
MethodUGAD: Universal Generative AI Detector utilizing Frequency Fingerprints
Alam, I., Muneer, M. S., & Woo, S. S. (2024) · ACM International Conference on Information and Knowledge Management (CIKM) · doi:10.1145/3627673.3680085 · Cited by 21
“As shown in the last row in Table 1, our method achieves the highest average AUC and accuracy, compared to the existing SOTA approaches: Wang et al., [63], Chandrasegaran et al., [7], Chai et al., [5], Grag et al., [15], and Xu Zhang et al. [71].”
Artificial Intelligence and the Spread of Mis- and Disinformation
Benzie, A., & Montasari, R. (2022) · Springer eBooks · doi:10.1007/978-3-031-06709-9_1 · Cited by 20
Perceptual Artifacts Localization for Inpainting
Zhang, L., Zhou, Y., Barnes, C., Amirghodsi, S., Lin, Z., Shechtman, E., & Shi, J. (2022) · Lecture notes in computer science · doi:10.1007/978-3-031-19818-2_9 · Cited by 13
TrueFace: a Dataset for the Detection of Synthetic Face Images from Social Networks
Boato, G., Pasquini, C., Stefani, A. L., Verde, S., & Miorandi, D. (2022) · IEEE International Joint Conference on Biometrics (IJCB) · doi:10.1109/ijcb54206.2022.10007988 · Cited by 15
“In [39], the authors observed the presence of spectral peaks caused by the upsampling operations performed in most GAN architectures.”
BackgroundDeepfake Video Detection Based on Convolutional Neural Networks
Adnan, S. R., & Abdulbaqi, H. A. (2022) · International Conference on Data Science and Intelligent Computing (ICDSIC) · doi:10.1109/icdsic56987.2022.10075830 · Cited by 14
CycleGAN without checkerboard artifacts for counter-forensics of fake-image detection
Osakabe, T., Tanaka, M., Kinoshita, Y., & Kiya, H. (2021) · International Workshop on Advanced Imaging Technology (IWAIT) 2021 · doi:10.1117/12.2590977 · Cited by 13
“Many researchers have been working on detecting images generated by using GANs.(8,9) In typical fake-image detection methods such as the method proposed by Zhang et al.”
BackgroundMethodInfluentialManipulated Face Detector: Joint Spatial and Frequency Domain Attention NetworkPreprint
Chen, Z., & Yang, H. (2020) · arXiv (Cornell University) · Cited by 13
SoK: Machine Learning GovernancePreprint
Chandrasekaran, V., Jia, H., Thudi, A., Travers, A., Yaghini, M., & Papernot, N. (2021) · arXiv (Cornell University) · doi:10.48550/arxiv.2109.10870 · Cited by 33
Faketracer: Exposing Deepfakes with Training Data Contamination
Sun, P., Li, Y., Qi, H., & Lyu, S. (2022) · Proceedings - International Conference on Image Processing · doi:10.1109/icip46576.2022.9897756 · Cited by 12
“Since the synthesized faces have highly natural looking and nonexistent identity in real-world, they can introduce misinformation and confuse the society [9].”
BackgroundDIFLD: domain invariant feature learning to detect low-quality compressed face forgery images
Zou, Y. L., Luo, C., & Zhang, J. (2023) · Complex & Intelligent Systems · doi:10.1007/s40747-023-01160-x · Cited by 14
A Robust Open-Set Multi-Instance Learning for Defending Adversarial Attacks in Digital Image
Uddin, K., Yang, Y., Jeong, T. H., & Oh, B. T. (2023) · IEEE Transactions on Information Forensics and Security · doi:10.1109/tifs.2023.3346211 · Cited by 12
“Zhang et al. [21] studied different GAN models and identified unique artifacts caused by upsam-pling.”
BackgroundMethodA Deepfake Image Detection Method Based on a Multi-Graph Attention Network
Chen, G., Du, C., Yuan, Y., Hu, H. G., Duan, H., & Zhu, H. (2025) · Electronics · doi:10.3390/electronics14030482 · Cited by 17
“Existing AI-generated image detectors, such as CNNSpot [32], Spec [33], and F3Net [34], demonstrate notable results.”
An Overview of Deepfake: The Sword of Damocles in AI
Tong, X., Wang, L., Pan, X., & Wang, J. G. (2020) · 2020 International Conference on Computer Vision, Image and Deep Learning (CVIDL) · doi:10.1109/cvidl51233.2020.00-88 · Cited by 12
A Comprehensive Review on Fake Images/Videos Detection Techniques
Chauhan, R., Popli, R., & Kansal, I. (2022) · International Conference on Reliability, Infocom Technologies and Optimization (Trends and Future Directions) (ICRITO) · doi:10.1109/icrito56286.2022.9964871 · Cited by 12
“Researchers in multimedia forensics experts have been working on this content for many times, presenting a variety of solutions to specific problems [7-9].”
BackgroundFDS_2D: rethinking magnitude-phase features for DeepFake detection
Yang, G., Wei, A., Fang, X., & Zhang, J. (2023) · Multimedia Systems · doi:10.1007/s00530-023-01118-6 · Cited by 13
Exposing GAN-Generated Profile Photos from Compact Embeddings
Mundra, S., Aniano, G., Marvaniya, S., Verbus, J. R., & Farid, H. (2023) · IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW) · doi:10.1109/cvprw59228.2023.00095 · Cited by 12
“Other artifacts include spatial frequency or noise anomalies [3, 9, 21, 33], but these artifacts tend to be vulnerable to simple laundering attacks (e.”
BackgroundLearning to Disentangle GAN Fingerprint for Fake Image AttributionPreprint
Yang, T., Cao, J., Sheng, Q., Li, L., Ji, J., Li, X., & Tang, S. (2021) · arXiv (Cornell University) · doi:10.48550/arxiv.2106.08749 · Cited by 15
“Some works [9, 10, 36] exploited the common checkerboard artifacts caused by upsampling operation in GAN architecture, and model this artifact in the frequency domain.”
BackgroundDarvish, M., Pouramini, M., & Bahador, H. J. (2022) · Iranian Conference on Machine Vision and Image Processing · doi:10.1109/mvip53647.2022.9738759 · Cited by 11
Single-Shot Focus Estimation for Microscopy Imaging With Kernel Distillation
Gu, Y. (2023) · IEEE Transactions on Computational Imaging · doi:10.1109/tci.2023.3281189 · Cited by 11
“These methods can easily over-fit to training samples and introduce the artefacts especially for methods based on GANs [32].”
MethodCGFormer: ViT-Based Network for Identifying Computer-Generated Images With Token Labeling
Quan, W., Deng, P., Wang, K., & Dong-ming Yan, A. (2023) · IEEE Transactions on Information Forensics and Security · doi:10.1109/tifs.2023.3322083 · Cited by 12
“2) Frequency Domain: Zhang et al. [39] proposed a spectrum-based classifier to identify GAN-generated images.”
BackgroundMethodRobust Adversarial Watermark Defending Against GAN Synthesization Attack
Xu, S., Qiao, T., Xu, M., Wang, W., & Zheng, N. (2024) · IEEE Signal Processing Letters · doi:10.1109/lsp.2024.3350983 · Cited by 11
Causal Representation Learning for GAN-Generated Face Image Quality Assessment
Tian, Y., Wang, S., Chen, B., & Kwong, S. (2024) · IEEE Transactions on Circuits and Systems for Video Technology · doi:10.1109/tcsvt.2024.3374887 · Cited by 11
“In addition, the distortions of GFIs differ from the traditional distortions and vary with network architectures, loss functions and processing tasks [22], [23], [24], [25].”
BackgroundFighting Malicious Media Data: A Survey on Tampering Detection and Deepfake Detection
Wang, J., Li, Z., Zhang, C., Chen, J., Wu, Z., Davis, L. S., & Jiang, Y. (2025) · Proceedings of the IEEE · doi:10.1109/jproc.2025.3576367 · Cited by 18
“Zhang et al. [299], Frank et al. [300], and Huang et al. [301] study the unique artifacts induced by the upsampling of GAN pipelines to develop robust GAN image classifiers.”
Robust Medical Image Colorization with Spatial Mask-Guided Generative Adversarial Network
Zhang, Z., Li, Y., & Shin, B. (2022) · Bioengineering · doi:10.3390/bioengineering9120721 · Cited by 13
“Although the aforementioned GAN-based methods have achieved promising results, these GAN-based methods mainly focus on adversarial training operated on the whole image while neglecting the various generated artifacts [17,18].”
BackgroundOpen Set Classification of GAN-based Image Manipulations via a ViT-based Hybrid Architecture
Wang, J., Alamayreh, O., Tondi, B., & Barni, M. (2023) · IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW) · doi:10.1109/cvprw59228.2023.00102 · Cited by 11
Frequency-Domain-Based Structure Losses for CycleGAN-Based Cone-Beam Computed Tomography Translation
Pai, S., Hadžić, I., Rao, C., Zhovannik, I., Dekker, A. L. A. J., Traverso, A., Asteriadis, S., & Hortal, E. (2023) · Sensors · doi:10.3390/s23031089 · Cited by 10
“Presence of artifacts or undesirable elements: The induction of artifacts is an established drawback of GAN-based generative models [30].”
BackgroundTowards Robust Gan-Generated Image Detection: A Multi-View Completion Representation
Liu, C., Zhu, T., Shen, S., & Zhou, W. (2023) · International Joint Conference on Artificial Intelligence · doi:10.24963/ijcai.2023/52 · Cited by 10
“[ Zhang et al. , 2019 ] pointed out that a detector can generalize between two GANs with similar spectral artifacts in their generated images, which in turn confirms that a detector is likely to overfit specific frequency patterns.”
BackgroundDisentangling different levels of GAN fingerprints for task-specific forensics
Liu, C., Zhu, T., Zhao, Y., Zhang, J., & Zhou, W. (2023) · Computer Standards & Interfaces · doi:10.1016/j.csi.2023.103825 · Cited by 10
Attacking Image Splicing Detection and Localization Algorithms Using Synthetic Traces
Fang, S., & Stamm, M. C. (2023) · IEEE Transactions on Information Forensics and Security · doi:10.1109/tifs.2023.3346312 · Cited by 12
“In order to make the generator capable of synthesizing forensic traces to avoid visible up-sampling artifacts [37], we use a fully convolutional architecture, where each convolutional layer uses a stride of 1.”
MethodData-agnostic Face Image Synthesis Detection using Bayesian CNNs
Leyva, R., Sánchez, V., Epiphaniou, G., & Maple, C. R. (2024) · Pattern Recognition Letters · doi:10.1016/j.patrec.2024.04.008 · Cited by 11
Xu, Q., Jiang, X., Sun, T., Wang, H., Meng, L., & Yan, H. (2024) · Information Fusion · doi:10.1016/j.inffus.2024.102578 · Cited by 11
Multi-Scale High-Frequency Vision Transformer for Face Forgery Detection
Chen, Z., Li, H., & Hu, J. (2025) · IEEE Transactions on Biometrics Behavior and Identity Science · doi:10.1109/tbiom.2025.3543692 · Cited by 9
“As described by Zhang et al. [19], most of the existing surface manipulation methods are based on GANs, where the upsampling operation results in abnormal frequency statistical properties of the forged surfaces.”
Evading Forensic Classifiers with Attribute-Conditioned Adversarial Faces
Shamshad, F., Srivatsan, K., & Nandakumar, K. (2023) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr52729.2023.01580 · Cited by 10
“The results reported in these works tend to indicate that simple, supervised deep learning-based classifiers are often highly effective at detecting GAN-generated images [35, 43].”
BackgroundEnhancing Interpretability in AI-Generated Image Detection with Genetic Programming
Lin, M., Shang, L., & Gao, X. (2023) · IEEE ... International Conference on Data Mining workshops · doi:10.1109/icdmw60847.2023.00053 · Cited by 10
“Some studies [24] [22] also pay attention to build a model with good generalization that can distinguish AI-generated images produced by unseen models.”
BackgroundAuthenticating AI-Generated Social Media Images Using Frequency Domain Analysis
Poredi, N., Nagothu, D., & Chen, Y. (2024) · IEEE Consumer Communications and Networking Conference · doi:10.1109/ccnc51664.2024.10454640 · Cited by 13
“Most GAN-based AI image generators showed distinct peaks in their mean DFTs [38], confirming the presence of patterns, which became the fingerprints that were used to detect them.”
BackgroundDoFNet: Depth of Field Difference Learning for Detecting Image Forgery
Jeong, Y., Choi, J., Kim, D., Park, S., Hong, M., Park, C., Min, S.-J., & Gwon, Y. (2021) · Lecture notes in computer science · doi:10.1007/978-3-030-69544-6_6 · Cited by 11
“…network [4,8,26–32] while the frequency-based method converts the pixel domain (i.e., 2D data) into the frequency domain to take the frequency spectrum as an input of the classification network [5–7, Another image forensic method is to detect the image areas manipulated by Adobe Photoshop.”
MethodEvading generated-image detectors: A deep dithering approach
Xie, H., Ni, J., Zhang, J., Zhang, W., & Huang, J. (2022) · Signal Processing · doi:10.1016/j.sigpro.2022.108558 · Cited by 11
Combining Identity Features and Artifact Analysis for Differential Morphing Attack Detection
Domenico, N. D., Borghi, G., Franco, A., & Maltoni, D. (2023) · Lecture notes in computer science · doi:10.1007/978-3-031-43148-7_9 · Cited by 10
Decentralized Attribution of Generative ModelsPreprint
Kim, C., Ren, Y. C., & Yang, Y. (2020) · arXiv (Cornell University) · doi:10.48550/arxiv.2010.13974 · Cited by 14
“In this context, it has been shown that generative models based on convolutional neural network leave fingerprints in their outputs (Odena et al. (2016)), allowing them to be detected (Zhang et al. (2019b); Marra et al. (2019); Wang et al. (2019a)).”
FakeTagger: Robust Safeguards against DeepFake Dissemination via Provenance TrackingPreprint
Wang, R., Juefei-Xu, F., Luo, M., Liu, Y., & Wang, L. (2020) · arXiv (Cornell University) · doi:10.48550/arxiv.2009.09869 · Cited by 10
“In the black-box setting, we employ a GAN simulator proposed in AutoGAN [65] to simulate the generation of DeepFake transformation.”
BackgroundMethodInfluentialWhodunit: Detection and Attribution of Synthetic Images by Leveraging Model-specific Fingerprints
Wißmann, A., Zeiler, S., Nickel, R. M., & Kolossa, D. (2024) · ACM International Workshop on Multimedia AI against Disinformation (MAD) · doi:10.1145/3643491.3660280 · Cited by 10
“Another common strand of detection methods uses handcrafted features [10, 11, 37, 42] that highlight unique characteristics of synthetic images, which are introduced by the respective generative model [25].”
BackgroundPandemia, polarización y odio: características de la desinformación en España
Blanco‐Herrero, D., Calderón, C. A., & Tovar, M. (2024) · Estudios sobre el Mensaje Periodístico · doi:10.5209/emp.96593 · Cited by 9
Exposing AI Generated Deepfake Images Using Siamese Network with Triplet Loss
Kanwal, S., Tehsin, S., & Saif, S. (2022) · Computing and Informatics · doi:10.31577/cai_2022_6_1541 · Cited by 10
The societal impact of Deepfakes: Advances in Detection and Mitigation
Kaushal, A., Mina, A., Meena, A., & Babu, T. H. (2023) · International Conference on Computing, Communication and Networking Technologies (ICCCNT) · doi:10.1109/icccnt56998.2023.10307353 · Cited by 11
“Generative adversarial networks (GANs): In paper [12] GANs are a type of neural network that can generate realistic images.”
BackgroundCapturing the Lighting Inconsistency for Deepfake Detection
Wu, W., Zhou, W., Zhang, W., Fang, H., & Yu, N. (2022) · Lecture notes in computer science · doi:10.1007/978-3-031-06788-4_52 · Cited by 9
Liu, S., Jiang, Q., Jin, X., He, Z., Zhou, W., Yao, S., & Wang, Q. (2022) · IEEE International Conference on Tools with Artificial Intelligence (ICTAI) · doi:10.1109/ictai56018.2022.00204 · Cited by 10
“…[29], methods fail to keep eyes on the complete frequency domain; b) after converting the entire images to the frequency domain via FFT or DFT, methods [30] [31] [32] directly put the feature map to classification, but manipulation clues are unnoticeable in the frequency domain of the whole image.”
MethodMining Generalized Features for Detecting AI-Manipulated Fake FacesPreprint
Yu, Y., Ni, R., & Zhao, Y. (2020) · arXiv (Cornell University) · doi:10.48550/arxiv.2010.14129 · Cited by 9
“Zhang et al .”
BackgroundInfluentialLEDNet: a multimodal foundation model for robust deepfake detection
Tao, R., Tang, S., Qin, H., Wang, W., Wei, Y., & Zhao, Y. (2025) · Science China Information Sciences · doi:10.1007/s11432-024-4400-8 · Cited by 9
“[16] simulated frequency artifacts to enable the detector to distinguish a deepfake of unknown source.”
General GAN-generated Image Detection by Data Augmentation in Fingerprint Domain
Wang, H., Fei, J., Dai, Y., Leng, L., & Xia, Z. (2023) · IEEE International Conference on Multimedia and Expo (ICME) · doi:10.1109/icme55011.2023.00207 · Cited by 8
“Most existing works [4]–[8] have achieved flawless detection accuracy on images from the same distribution as the training data, i.”
BackgroundCharacterizing Photorealism and Artifacts in Diffusion Model-Generated Images
Kamali, N., Nakamura, K., Kumar, A., Chatzimparmpas, A., Hullman, J. R., & Groh, M. (2025) · CHI Conference on Human Factors in Computing Systems · doi:10.1145/3706598.3713962 · Cited by 23
“…from 50,444 participants on 599 images to enable the replications of our results and further research on photorealism in diffusion models; replication data, code, and stimuli can be found at the following link: https://github.com/negarkamali/Replication-for-Characterizing-Photorealism-2025/.”
Training humans for synthetic face image detection
Rehman, R. S., Meier, E., Ibsen, M., Rathgeb, C., Nichols, R., & Busch, C. (2025) · Frontiers in Artificial Intelligence · doi:10.3389/frai.2025.1568267 · Cited by 8
Noise-Informed Diffusion-Generated Image Detection With Anomaly Attention
Guan, W., Wang, W., Peng, B., He, Z., Dong, J., & Cheng, H. (2025) · IEEE Transactions on Information Forensics and Security · doi:10.1109/tifs.2025.3573161 · Cited by 8
“Previous forgery detection methods are primarily designed to detect GAN-generated (Generative Adversarial Networks [12]) images [13], [14], [15].”
Adversarial Attack on Fake-Faces Detectors Under White and Black Box Scenarios
Wang, X., Ni, R., Li, W., & Zhao, Y. (2021) · Proceedings - International Conference on Image Processing · doi:10.1109/icip42928.2021.9506273 · Cited by 8
“Early methods are based on color saturation differences [4], spectral artifacts [5], GAN fingerprint [6] etc, which have been”
MethodHierarchical Learning and Dummy Triplet Loss for Efficient Deepfake Detection
Beuve, N., Hamidouche, W., & Olivier Déforges, A. (2023) · ACM Transactions on Multimedia Computing Communications and Applications · doi:10.1145/3626101 · Cited by 8
“They may identify artifacts in the spatial domain, left by the wrapping of the target face on the source image [31], artifacts left by GANs [52] or by convolutions [21].”
BackgroundFinding AI-Generated Faces in the Wild
Aniano, G., Gindi, J., Mundra, S., Verbus, J. R., & Farid, H. (2024) · IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW) · doi:10.1109/cvprw63382.2024.00433 · Cited by 13
“Other non-face specific artifacts include spatial frequency or noise anomalies [5,8,12,22,36], but these artifacts tend to be vulnerable to simple laundering attacks (e.g., transcoding, additive noise, image resizing).”
BackgroundOn the Robustness and Generalizability of Face Synthesis Detection Methods
Sabel, J., & Johansson, F. (2021) · IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW) · doi:10.1109/cvprw53098.2021.00107 · Cited by 9
“[29] improve the generalization capability of a binary classifier (ResNet-34 [7] pre-trained on ImageNet) by using the image frequency spectrum as input instead of RGB pixels.”
BackgroundDual-Tree Complex Wavelet Transform-Based Direction Correlation for Face Forgery Detection
Gao, S., Xia, M., & Yang, G. (2021) · Security and Communication Networks · doi:10.1155/2021/8661083 · Cited by 9
“exploited 2D DCTfor each RGB channel of the original image and then used AutoGAN [38] to synthesize GAN artifacts in any image without pretrained model [15].”
MethodCross-Domain Local Characteristic Enhanced Deepfake Video Detection
Liu, Z., Wang, H., & Wang, S. (2023) · Lecture notes in computer science · doi:10.1007/978-3-031-26348-4_12 · Cited by 9
DiffGAR: Model-Agnostic Restoration from Generative Artifacts Using Image-to-Image Diffusion Models
Yin, Y., Huang, L., Liu, Y., & Huang, K. (2022) · doi:10.1145/3577530.3577539 · Cited by 9
“However, artifacts are often generated due to architectural limitations of the generative model itself and its inability to capture the complete data manifold pattern [38], [14], [32], highlighting the importance of generative artifacts removal for producing more visually appealing images.”
BackgroundRobust Attentive Deep Neural Network for Exposing GAN-generated FacesPreprint
Hui, G., Hu, S., Wang, X., Chang, M., & Lyu, S. (2021) · arXiv (Cornell University) · doi:10.48550/arxiv.2109.02167 · Cited by 8
“Many works [36]–[38] identify GAN images by recognizing the specific artifacts produced by the GAN upsampling process.”
BackgroundIs Training Useful to Detect Deepfakes? : A Preliminary Study.
Mohamed, N. M. A. A., Bogdanel, G., & Gómez-Moreno, H. (2023) · Iberian Conference on Information Systems and Technologies (CISTI) · doi:10.23919/cisti58278.2023.10211622 · Cited by 7
“These detection approaches operate in the spatial domain or even in the frequency domain [14, 15].”
BackgroundRana, M. S., & Sung, A. H. (2024) · International Conference on Information and Computer Technologies (ICICT) · doi:10.1109/icict62343.2024.00019 · Cited by 8
“Zhang et al. [17] introduced a GAN simulator that replicates collective GAN-generated image artifacts and inputs them into a classifier to identify Deepfakes.”
BackgroundManisha, M., Li, C., & Kotegar, K. A. (2025) · IEEE Access · doi:10.1109/access.2025.3537558 · Cited by 8
Revisiting face forgery detection towards generalization
Peng, C., Chen, T., Liu, D., Guo, H., Wang, N., & Gao, X. (2025) · Neural Networks · doi:10.1016/j.neunet.2025.107310 · Cited by 7
DiffFake: Exposing Deepfakes Using Differential Anomaly Detection
Stamnas, S., & Sánchez, V. (2025) · IEEE/CVF Winter Conference on Applications of Computer Vision Workshops (WACVW) · doi:10.1109/wacvw65960.2025.00079 · Cited by 8
“Furthermore, several methods have focused on extracting information from the frequency domain [20, 37, 38, 42, 59], which has also been shown to improve generalization capabilities.”
Nguyen, T. D., Azizpour, A., & Stamm, M. C. (2025) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr52734.2025.00289 · Cited by 12
“This holds true for both cameras and AI image generators [45, 73].”
SpecXNet: A Dual-Domain Convolutional Network for Robust Deepfake Detection
Alam, I., Islam, M. T., & Woo, S. S. (2025) · ACM International Conference on Multimedia (ACM MM) · doi:10.1145/3746027.3755707 · Cited by 15
Ma, H., Zhang, J., Chen, X., Chu, W., Guo, J., Zheng, J., Yang, L., & Liang, Y. (2025) · IET Biometrics · doi:10.1049/bme2/3764746 · Cited by 7
“In [20], the authors observed that the checkerboard artifacts always exist in frequency domain for face images generated by GAN, a GAN simulator is then designed to reveal these artifacts.”
Channel-Wise Spatiotemporal Aggregation Technology for Face Video Forensics
Lu, Y., Liu, Y., Fei, J., & Xia, Z. (2021) · Security and Communication Networks · doi:10.1155/2021/5524930 · Cited by 7
An evolutionary- convolutional neural network for fake image detection
Jasim, R. M., & Atia, T. S. (2022) · Indonesian Journal of Electrical Engineering and Computer Science · doi:10.11591/ijeecs.v29.i3.pp1657-1667 · Cited by 12
“It has been stated that face photos with varied qualities produced by the same GAN have the same artifacts or fingerprints [34].”
Reliable Out-of-Distribution Recognition of Synthetic Images
Maier, A., & Rieß, C. (2024) · Journal of Imaging · doi:10.3390/jimaging10050110 · Cited by 8
“In this more challenging scenario, Spec, PatchForensics, and Wang et al. drop to random guessing for all the generators [15,52,53].”
BackgroundMethodDeepGuard: Identification and Attribution of AI-Generated Synthetic Images
Namani, Y., Reghioua, I., Bendiab, G., Labiod, M. A., & Shiaeles, S. N. (2025) · Electronics · doi:10.3390/electronics14040665 · Cited by 10
FairAdapter: Detecting AI-generated Images with Improved Fairness
Ding, F., Zhang, J., He, X., & Xu, J. (2025) · IEEE International Conference on Acoustics Speech and Signal Processing · doi:10.1109/icassp49660.2025.10890705 · Cited by 8
“Zhang et al. [11] disclosed the artifacts induced during the up-sampling of the auto-encoder in the frequency domain.”
Not with My Name! Inferring Artists’ Names of Input Strings Employed by Diffusion Models
Leotta, R., Giudice, O., Guarnera, L., & Battiato, S. (2023) · Lecture notes in computer science · doi:10.1007/978-3-031-43148-7_31 · Cited by 7
D4: Detection of Adversarial Diffusion Deepfakes Using Disjoint Ensembles
Hooda, A., Mangaokar, N., Feng, R., Fawaz, K., Jha, S., & Prakash, A. (2024) · 2024 IEEE/CVF Winter Conference on Applications of Computer Vision (WACV) · doi:10.1109/wacv57701.2024.00377 · Cited by 7
“Similar work has also achieved remarkable performance using frequency features — [41, 42], and more recent efforts have emphasized their utility in detecting deepfakes from both GANs and diffusion models [43].”
BackgroundSeam Carving Detection and Localization Using Two-Stage Deep Neural Networks
Nataraj, L., Gudavalli, C., Mohammed, T. M., Chandrasekaran, S., & Manjunath, B. S. (2021) · Lecture notes in electrical engineering · doi:10.1007/978-981-16-0289-4_29 · Cited by 7
“Other approaches exploit JPEG compression artifacts [13,24,28,7] or artifacts arising from artificial intelligence (AI) generated images [30,48,16,3].”
Thinking in Patch: Towards Generalizable Forgery Detection with Patch Transformation
Zhang, X., Wang, S., Liu, C., Zhang, M., Liu, X., & Xie, H. (2021) · Lecture notes in computer science · doi:10.1007/978-3-030-89370-5_25 · Cited by 7
Self-citationLeveraging High-Frequency Components for Deepfake Detection
Mejri, N., Papadopoulos, K., & Aouada, D. (2021) · IEEE International Workshop on Multimedia Signal Processing (MMSP) · doi:10.1109/mmsp53017.2021.9733606 · Cited by 7
Making Generated Images Hard to Spot: A Transferable Attack on Synthetic Image Detectors
Zhao, X., & Stamm, M. C. (2023) · Lecture notes in computer science · doi:10.1007/978-3-031-37745-7_6 · Cited by 10
“techniques to detect GAN-generated images [9], [10], [11], [12], [13] and to attribute them to the specific GAN used to create them [14], [15], [16].”
BackgroundMethodSelf-supervised scheme for generalizing GAN image detection
Jeong, Y., Kim, D., Kim, P., Ro, Y., & Choi, J. (2024) · Pattern Recognition Letters · doi:10.1016/j.patrec.2024.06.030 · Cited by 6
Semantic Token Transformer for Face Forgery Detection
Peng, C., Luo, X., Liu, D., Wang, N., Hu, R., & Gao, X. (2025) · IEEE Transactions on Information Forensics and Security · doi:10.1109/tifs.2025.3567110 · Cited by 7
“For instance, some studies [41], [42], [43] use Discrete Fourier Transform to transform images from spatial to the frequency domain.”
Are CLIP Features All You Need for Universal Synthetic Image Origin Attribution?
Cioni, D., Tzelepis, C., Seidenari, L., & Patras, I. (2025) · Lecture notes in computer science · doi:10.1007/978-3-031-92648-8_22 · Cited by 15
“Further research in GAN attribution [22,25,73] validated the existence of such fingerprints, allowing to identify the generator among a finite, fixed (i.e., closed) set of models.”
BackgroundYumlembam, R., Issac, B., Aslam, N., Babu, E. K., Collyer, J., & Kennedy, F. (2025) · Scientific Reports · doi:10.1038/s41598-025-28572-8 · Cited by 7
“However, it has been shown in previous work that they do not generalise well and failed to detect AI-generated images when the AI image generation method changes 4 , 5 , 6 , 7 .”
Metadata
- Authors
- Zhang, X., Karaman, S., & Chang, S.-F.
- Affiliations
- Columbia University (US)
- Venue
- 2019 IEEE International Workshop on Information Forensics and Security (WIFS)
- Publisher
- IEEE
- Published
- 2019-12
- Pages
- 1-6
- ISSN
- 2157-4774 (ISSN-L)
- OpenAlex
- W3012472557
- Cited by
- 567
- References
- 21
- Citations by year
- 2026: 60 · 2025: 113 · 2024: 89 · 2023: 100 · 2022: 84 · 2021: 81 · 2020: 39
- FWCI
- 16.52
- Type
- conference paper
- Language
- en
- Open access
- green
- arxiv.org
- Licence
- Locations
- IEEE International Workshop on Information Forensics and Security (publishedVersion)
- arXiv (Cornell University) (submittedVersion)
- arXiv (Cornell University)
- Topics
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Digital Media Forensic Detection
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Generative Adversarial Networks and Image Synthesis
- Physical Sciences › Computer Science › Artificial Intelligence › Adversarial Robustness in Machine Learning
- Keywords
- GAN-generated image detection, CycleGAN, adversarial image generation
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
