DeepfakeGenome: Toward Next-Generation Deepfake Attribution
Organizations: Tencent Youtu Lab · Shanghai Jiao Tong University · East China Normal University · Peking University
Abstract
In recent years, AIGC technologies are capable of generating hyper-realistic forged facial images, posing severe threats to facial security safety. To tackle these challenges, early Deepfake research primarily centered on real-fake classification. Recently, growing efforts have been devoted to the Deepfake Attribution (DFA) of generated content. Nevertheless, existing research on Deepfake detection and attribution faces saturated performance in binary classification, limited diversity in datasets and algorithms, and imperfect evaluation protocols, which severely impede practical application. To address these limitations, we propose a comprehensive deepfake detection and attribution benchmark named DeepfakeGenome (DFG). It contains 100 facial forgery algorithms and 2M images in total, achieving 4x to 100x larger than prior DFA benchmarks. We further design 4 protocols for practical evaluation, including a novel retrieval-based attribution paradigm. Unlike previous open-set evaluation metrics, the proposed retrieval metrics are more aligned with the real-world active defense situation of blacklist registration mechanisms. Based on these elaborate designs, we investigate the performance ceiling of deepfake attribution task. Over 2k experimental evaluations are conducted, and 10 insightful findings are derived. We believe this work can provide new insights into the DFA research field. The Code and Benchmark of DFG are avaliable on https://huggingface.co/spaces/DeepfakeGenome/DFG_Page.
Figures & tables
| Types | Basic Backbone | Finetuned CLIP | Reconstruction-based | Frequency-domain | Latent Space | Representation | Aug Adv. | ||||||||
| Xception | Eff-B4 | CLIP-ViT-FFT | Effort | RECCE | SBI | F3Net | SRM | SPSL | ProDet | LSDA | UCF | IID | CORE | sladd | |
| [ 60 ] | [ 61 ] | [ 27 ] | [ 3 ] | [ 62 ] | [ 30 ] | [ 63 ] | [ 64 ] | [ 65 ] | [ 66 ] | [ 67 ] | [ 68 ] | [ 69 ] | [ 70 ] | [ 71 ] | |
| FR (3 Algos) | 0.6854 | 0.7144 | 0.7323 | 0.8252 | 0.6976 | 0.6937 | 0.7073 | 0.6644 | 0.6157 | 0.6801 | 0.7189 | 0.7157 | 0.6767 | 0.7101 | 0.7033 |
| FS (6 Algos) | 0.6416 | 0.6637 | 0.7752 | 0.8184 | 0.6194 | 0.6098 | 0.6088 | 0.6121 | 0.6465 | 0.6167 | 0.6128 | 0.6488 | 0.6087 | 0.6393 | 0.6057 |
| FE (11 Algos) | 0.5613 | 0.5939 | 0.6315 | 0.7586 | 0.5762 | 0.5897 | 0.5908 | 0.5846 | 0.5643 | 0.5603 | 0.5908 | 0.5908 | 0.6080 | 0.6032 | 0.6051 |
| EFS (39 Algos) | 0.5923 | 0.5486 | 0.5664 | 0.5895 | 0.6052 | 0.6123 | 0.5682 | 0.5821 | 0.5445 | 0.5021 | 0.6841 | 0.6841 | 0.6487 | 0.6398 | 0.5962 |
| Architecture | Basic Backbone | Finetuned CLIP | Reconstruction-based | Frequency-domain | Latent Space | Representation | Aug Adv. | ||||||||
| Xception | Eff-B4 | CLIP-ViT-FFT | Effort | RECCE | SBI | F3Net | SRM | SPSL | ProDet | LSDA | UCF | IID | CORE | sladd | |
| [ 60 ] | [ 61 ] | [ 27 ] | [ 3 ] | [ 62 ] | [ 30 ] | [ 63 ] | [ 64 ] | [ 65 ] | [ 66 ] | [ 67 ] | [ 68 ] | [ 69 ] | [ 70 ] | [ 71 ] | |
| CG (1 Algo) | 0.5252 | 0.5430 | 0.5837 | 0.5824 | 0.4954 | 0.4823 | 0.5072 | 0.5006 | 0.5359 | 0.4819 | 0.4863 | 0.5064 | 0.4751 | 0.5206 | 0.4821 |
| GAN (16 Algos) | 0.5975 | 0.6630 | 0.7352 | 0.8477 | 0.6157 | 0.6135 | 0.6430 | 0.5806 | 0.5894 | 0.5608 | 0.6395 | 0.6607 | 0.6122 | 0.6366 | 0.5923 |
| Diffusion (29 Algos) | 0.5874 | 0.5396 | 0.5710 | 0.6141 | 0.5926 | 0.4309 | 0.5865 | 0.5932 | 0.5665 | 0.5556 | 0.6440 | 0.5568 | 0.6479 | 0.6339 | 0.6114 |
| Close (11 Algos) | 0.6161 | 0.5508 | 0.5393 | 0.5153 | 0.6277 | 0.3876 | 0.6433 | 0.5983 | 0.5169 | 0.4464 | 0.7383 | 0.5757 | 0.6573 | 0.6509 | 0.6146 |
| Methods | Ret-mAP | Ret-Recall | Ret-Recall | Avg |
| @1% | @0.1% | |||
| Visual BackBones | ||||
| Xception [ 60 ] | 0.2493 | 0.4586 | 0.2916 | 0.3331 |
| F3Net [ 63 ] | 0.2736 | 0.4289 | 0.2567 | 0.3197 |
| Resnet-34 [ 72 ] | 0.2745 | 0.4242 | 0.276 | 0.3249 |
| Effort [ 3 ] | 0.5753 | 0.8607 | 0.7727 | 0.7362 |
| Methods | Protocol-2 | Protocol-3 | ||||||||||
| Acc | mAP | Ret-mAP | Ret-Recall | Ret-Recall | Avg | Acc | mAP | Ret-mAP | Ret-Recall | Ret-Recall | Avg | |
| @1% | @0.1% | @1% | @0.1% | |||||||||
| Visual BackBones | ||||||||||||
| Xception [ 60 ] | 0.8899 | 0.9176 | 0.9149 | 0.9515 | 0.8762 | 0.910 | 0.7997 | 0.8501 | 0.8519 | 0.8316 | 0.6824 | 0.8031 |
| F3Net [ 63 ] | 0.8884 | 0.9130 | 0.9106 | 0.9525 | 0.8904 | 0.9110 | 0.8220 | 0.8606 | 0.8395 | 0.8433 | 0.7177 | 0.8166 |
| ResNet-34 [ 72 ] | 0.8755 | 0.9044 | 0.8899 | 0.9426 | 0.8923 | 0.9009 | 0.7446 | 0.8116 | 0.8158 | 0.8170 | 0.6140 | 0.7606 |
| Methods | Process | Protocol-2 | Protocol-3 | ||||||||||
| Acc | mAP | Ret-mAP | Ret-Recall | Ret-Recall | Avg | Acc | mAP | Ret-mAP | Ret-Recall | Ret-Recall | Avg | ||
| @1% | @0.1% | @1% | @0.1% | ||||||||||
| CLIP-ViT | - | 0.9021 | 0.9334 | 0.9128 | 0.9786 | 0.9339 | 0.9322 | 0.8091 | 0.8711 | 0.8776 | 0.9294 | 0.8597 | 0.8694 |
| Vanilla-ViT | - | 0.8667 | 0.9038 | 0.8749 | 0.9290 | 0.8254 | 0.8799 | 0.7367 | 0.8099 | 0.8163 | 0.8087 | 0.6454 | 0.7634 |
| CLIP-ViT | w. CL | 0.9023 | 0.9320 | 0.9191 | 0.9703 | 0.9220 | 0.9291 | 0.8193 | 0.8674 | 0.8800 | 0.8965 | 0.7690 | 0.8464 |
| CLIP-ViT | w. LSDA [ 67 ] | 0.8975 | 0.9377 | 0.9107 | 0.9748 | 0.9391 | 0.9320 | 0.8048 | 0.8587 | 0.8674 | 0.9213 | 0.8144 | 0.8533 |
Appendix figures & tables10 assets
Supplementary material from the paper’s appendix.
Appendix
| Dataset | Task | ID | Algorithm | Venue | Type | Domain | FF++ Train | FF++ Test | Other Real | Other Fake |
| FF++ [ 13 ] | FS | 001 | Deepfakes [ 55 ] | None | GAN | FF++ | 22928 | 4454 | - | - |
| FS | 002 | FaceShifter [ 80 ] | ArXiv 2019 | CNN | FF++ | 22968 | 4479 | - | - | |
| FR | 003 | NeuralTextures [ 58 ] | TOG 2019 | GAN | FF++ | 22973 | 4479 | - | - | |
| FR | 004 | Face2Face [ 56 ] | CVPR 2016 | CG | FF++ | 22991 | 4480 | - | - | |
| DF40 [ 4 ] | FS | 005 | FSGAN [ 81 ] | ICCV 2019 | GAN | FF++ & CDF | 21815 | 4455 | CDF-5.5k | 19601 |
| FS | 006 | SimSwap [ 82 ] | MM 2020 | GAN | FF++ & CDF | 31350 | 4468 | CDF-5.5k | 20322 |
| Training Set | Algorithms |
| DeepFakes [ 55 ] , Face2Face [ 56 ] , FaceSwap [ 57 ] , NeuralTextures [ 58 ] | ATVG-Net [ 117 ] , FirstOrderMotion [ 118 ] , Talking-headVideo [ 119 ] , DiscoFaceGAN [ 120 ] , MaskGAN [ 121 ] , SC-FEGAN [ 122 ] , MMreplacement [ 14 ] , DSS [ 14 ] , SBS [ 14 ] , FaceAPP [ 123 ] , StyleGAN [ 124 ] , ProGAN [ 125 ] , MMDGAN [ 126 ] , SNGAN [ 127 ] , CramerGAN [ 128 ] , InfoMax-GAN [ 129 ] , SSGAN [ 130 ] , AttGAN [ 131 ] , Adobe Firefly [ 132 ] , DALL·E3 [ 133 ] , Freepik [ 134 ] , Leonardo AI [ 135 ] , Starry AI [ 136 ] , DALL·E 1 [ 137 ] , Deep AI [ 138 ] , Hotpot AI [ 139 ] , FLUX.1 [ 140 ] , FLUX 1.1 Pro [ 141 ] , SD 3.5 Large [ 142 ] , SD XL Turbo [ 142 ] , NVIDIA Sana PAG [ 143 ] , Stable Cascade [ 144 ] , SD Attend&Excite [ 145 ] , Hunyuan-DiT [ 146 ] , Midjourney [ 147 ] , SD XL [ 148 ] , FreeDoM_T [ 149 ] , HPS [ 150 ] , SDXL Refiner [ 148 ] , DreamBooth [ 151 ] , DiffFace [ 152 ] , DCFace [ 153 ] , Imagic [ 154 ] , CollabDiff [ 155 ] , CycleDiff [ 156 ] , PNDM [ 157 ] , P2/ADM [ 158 ] , LDM [ 142 ] , SD v1.5 [ 142 ] , SDEdit [ 159 ] , DiffSwap [ 160 ] , Qwen-Image [ 1 ] , BAGEL [ 20 ] , HunyuanImage-2.1 [ 21 ] , HunyuanImage-3.0 [ 22 ] , Infinity [ 23 ] , Hart [ 24 ] , Nano Banana [ 2 ] , GPT-4o [ 25 ] |
| Training Set | Test Set |
| Deepfakes [ 55 ] , FaceShifter [ 80 ] , NeuralTextures [ 58 ] , Face2Face [ 56 ] , FSGAN [ 81 ] , SimSwap [ 82 ] , BlendFace [ 83 ] , e4s [ 84 ] , FaceDancer [ 85 ] , MobileSwap [ 86 ] , InSwapper [ 87 ] , UniFace [ 88 ] , FaceSwap [ 57 ] , Sadtalker [ 90 ] , MRAA [ 91 ] , LIA [ 92 ] , FS_vid2vid [ 93 ] , Wav2Lip [ 94 ] , DaGAN [ 95 ] , Hyperreenact [ 96 ] , MCNet [ 97 ] , OneShot [ 98 ] , PIRender [ 99 ] , TPSMM [ 100 ] , FOMM [ 101 ] , HeyGen [ 102 ] , VQGAN [ 103 ] , StyleGAN3 [ 104 ] , StyleGAN-XL [ 105 ] , DDPM [ 107 ] , SD-2.1 [ 108 ] , RDDM [ 109 ] , PixArt- [ 110 ] , DiT-XL/2 [ 111 ] , SiT-XL/2 [ 112 ] , e4e [ 113 ] | Deepfakes [ 55 ] , FaceShifter [ 80 ] , NeuralTextures [ 58 ] , Face2Face [ 56 ] , FSGAN [ 81 ] , SimSwap [ 82 ] , BlendFace [ 83 ] , e4s [ 84 ] , FaceDancer [ 85 ] , MobileSwap [ 86 ] , InSwapper [ 87 ] , UniFace [ 88 ] , FaceSwap [ 57 ] , Sadtalker [ 90 ] , MRAA [ 91 ] , LIA [ 92 ] , FS_vid2vid [ 93 ] , Wav2Lip [ 94 ] , DaGAN [ 95 ] , Hyperreenact [ 96 ] , MCNet [ 97 ] , OneShot [ 98 ] , PIRender [ 99 ] , TPSMM [ 100 ] , FOMM [ 101 ] , HeyGen [ 102 ] , VQGAN [ 103 ] , StyleGAN3 [ 104 ] , StyleGAN-XL [ 105 ] , DDPM [ 107 ] , SD-2.1 [ 108 ] , RDDM [ 109 ] , PixArt- [ 110 ] , DiT-XL/2 [ 111 ] , SiT-XL/2 [ 112 ] , e4e [ 113 ] |
| Training Set | Testing Set |
| Deepfakes [ 55 ] , FaceShifter [ 80 ] , NeuralTextures [ 58 ] , Face2Face [ 56 ] , FSGAN [ 81 ] , SimSwap [ 82 ] , BlendFace [ 83 ] , e4s [ 84 ] , FaceDancer [ 85 ] , MobileSwap [ 86 ] , InSwapper [ 87 ] , UniFace [ 88 ] , FaceSwap [ 57 ] , Sadtalker [ 90 ] , MRAA [ 91 ] , LIA [ 92 ] , FS_vid2vid [ 93 ] , Wav2Lip [ 94 ] , DaGAN [ 95 ] , Hyperreenact [ 96 ] , MCNet [ 97 ] , OneShot [ 98 ] , PIRender [ 99 ] , TPSMM [ 100 ] , FOMM [ 101 ] , HeyGen [ 102 ] , VQGAN [ 103 ] , StyleGAN3 [ 104 ] , StyleGAN-XL [ 105 ] , DDPM [ 107 ] , SD-2.1 [ 108 ] , RDDM [ 109 ] , PixArt- [ 110 ] , DiT-XL/2 [ 111 ] , SiT-XL/2 [ 112 ] , e4e [ 113 ] | FSGAN [ 81 ] , SimSwap [ 82 ] , BlendFace [ 83 ] , e4s [ 84 ] , FaceDancer [ 85 ] , MobileSwap [ 86 ] , InSwapper [ 87 ] , UniFace [ 88 ] , FaceSwap [ 57 ] , Sadtalker [ 90 ] , MRAA [ 91 ] , LIA [ 92 ] , FS_vid2vid [ 93 ] , Wav2Lip [ 94 ] , DaGAN [ 95 ] , Hyperreenact [ 96 ] , MCNet [ 97 ] , OneShot [ 98 ] , PIRender [ 99 ] , TPSMM [ 100 ] , FOMM [ 101 ] , HeyGen [ 102 ] , VQGAN [ 103 ] , StyleGAN3 [ 104 ] , StyleGAN-XL [ 105 ] , DDPM [ 107 ] , SD-2.1 [ 108 ] , RDDM [ 109 ] , PixArt- [ 110 ] , DiT-XL/2 [ 111 ] , SiT-XL/2 [ 112 ] , e4e [ 113 ] |
| Training Set | Testing Set |
| Deepfakes [ 55 ] , FaceShifter [ 80 ] , NeuralTextures [ 58 ] , Face2Face [ 56 ] , FSGAN [ 81 ] , SimSwap [ 82 ] , BlendFace [ 83 ] , e4s [ 84 ] , FaceDancer [ 85 ] , MobileSwap [ 86 ] , InSwapper [ 87 ] , UniFace [ 88 ] , FaceSwap [ 57 ] , Sadtalker [ 90 ] , MRAA [ 91 ] , LIA [ 92 ] , FS_vid2vid [ 93 ] , Wav2Lip [ 94 ] , DaGAN [ 95 ] , Hyperreenact [ 96 ] , MCNet [ 97 ] , OneShot [ 98 ] , PIRender [ 99 ] , TPSMM [ 100 ] , FOMM [ 101 ] , HeyGen [ 102 ] , VQGAN [ 103 ] , StyleGAN3 [ 104 ] , StyleGAN-XL [ 105 ] , DDPM [ 107 ] , SD-2.1 [ 108 ] , RDDM [ 109 ] , PixArt- [ 110 ] , DiT-XL/2 [ 111 ] , SiT-XL/2 [ 112 ] , e4e [ 113 ] | ATVG-Net [ 117 ] , FirstOrderMotion [ 118 ] , Talking-headVideo [ 119 ] , DiscoFaceGAN [ 120 ] , MaskGAN [ 121 ] , SC-FEGAN [ 122 ] , MMreplacement [ 14 ] , DSS [ 14 ] , SBS [ 14 ] , FaceAPP [ 123 ] , StyleGAN [ 124 ] , ProGAN [ 125 ] , MMDGAN [ 126 ] , SNGAN [ 127 ] , CramerGAN [ 128 ] , InfoMax-GAN [ 129 ] , SSGAN [ 130 ] , AttGAN [ 131 ] , Adobe Firefly [ 132 ] , DALL·E3 [ 133 ] , Freepik [ 134 ] , Leonardo AI [ 135 ] , Starry AI [ 136 ] , DALL·E 1 [ 137 ] , Deep AI [ 138 ] , Hotpot AI [ 139 ] , FLUX.1 [ 140 ] , FLUX 1.1 Pro [ 141 ] , SD 3.5 Large [ 142 ] , SD XL Turbo [ 142 ] , NVIDIA Sana PAG [ 143 ] , Stable Cascade [ 144 ] , SD Attend&Excite [ 145 ] , Hunyuan-DiT [ 146 ] , Midjourney [ 147 ] , SD XL [ 148 ] , FreeDoM_T [ 149 ] , HPS [ 150 ] , SDXL Refiner [ 148 ] , DreamBooth [ 151 ] , DiffFace [ 152 ] , DCFace [ 153 ] , Imagic [ 154 ] , CollabDiff [ 155 ] , CycleDiff [ 156 ] , PNDM [ 157 ] , P2/ADM [ 158 ] , LDM [ 142 ] , SD v1.5 [ 142 ] , SDEdit [ 159 ] , DiffSwap [ 160 ] , Qwen-Image [ 1 ] , BAGEL [ 20 ] , HunyuanImage-2.1 [ 21 ] , HunyuanImage-3.0 [ 22 ] , Infinity [ 23 ] , Hart [ 24 ] , Nano Banana [ 2 ] , GPT-4o [ 25 ] |
| Algorithm | Basic Backbone | Finetuned CLIP | Reconstruction-based | Frequency-domain | Latent Space | Representation | Aug Adv. | ||||||||
| Xception | Eff-B4 | CLIP-L-FFT | Effort | RECCE | SBI | F3Net | SRM | SPSL | ProDet | LSDA | UCF | IID | CORE | sladd | |
| [ 60 ] | [ 61 ] | [ 27 ] | [ 3 ] | [ 62 ] | [ 30 ] | [ 63 ] | [ 64 ] | [ 65 ] | [ 66 ] | [ 67 ] | [ 68 ] | [ 69 ] | [ 70 ] | [ 71 ] | |
| ATVG-Net [ 117 ] | 0.7110 | 0.7480 | 0.7601 | 0.8360 | 0.7412 | 0.8347 | 0.7588 | 0.7037 | 0.5620 | 0.6876 | 0.7553 | 0.7827 | 0.7287 | 0.7453 | 0.7494 |
| FirstOrderMotion [ 118 ] | 0.7910 | 0.7955 | 0.8190 | 0.8679 | 0.7922 | 0.6973 | 0.8116 | 0.7681 | 0.6698 | 0.7347 | 0.8019 | 0.8256 | 0.7811 | 0.8069 | 0.7743 |
| Talking-headVideo [ 119 ] | 0.5541 | 0.5997 | 0.6177 | 0.7717 | 0.5593 | 0.5491 | 0.5514 | 0.5214 | 0.6152 | 0.6181 | 0.5996 | 0.5390 | 0.5203 | 0.5782 | 0.5863 |
| MMreplacement [ 14 ] | 0.5252 | 0.5430 | 0.5837 | 0.5824 | 0.4954 | 0.4823 | 0.5072 | 0.5006 | 0.5359 | 0.4819 | 0.4863 | 0.5064 | 0.4751 | 0.5206 | 0.4821 |
| Algorithm | Basic Backbone | Finetuned CLIP | Reconstruction-based | Frequency-domain | Latent Space | Representation | Aug Adv. | ||||||||
| Xception | Eff-B4 | CLIP-L-FFT | Effort | RECCE | SBI | F3Net | SRM | SPSL | ProDet | LSDA | UCF | IID | CORE | sladd | |
| [ 60 ] | [ 61 ] | [ 27 ] | [ 3 ] | [ 62 ] | [ 30 ] | [ 63 ] | [ 64 ] | [ 65 ] | [ 66 ] | [ 67 ] | [ 68 ] | [ 69 ] | [ 70 ] | [ 71 ] | |
| ATVG-Net [ 117 ] | 0.8039 | 0.8392 | 0.9452 | 0.9808 | 0.8267 | 0.8875 | 0.8246 | 0.8346 | 0.7674 | 0.9519 | 0.8559 | 0.8725 | 0.8331 | 0.8392 | 0.8996 |
| FirstOrderMotion [ 118 ] | 0.9016 | 0.8880 | 0.9734 | 0.9824 | 0.8925 | 0.7724 | 0.9000 | 0.8911 | 0.8559 | 0.8773 | 0.9034 | 0.9184 | 0.8844 | 0.9082 | 0.8910 |
| Talking-headVideo [ 119 ] | 0.5984 | 0.6486 | 0.6682 | 0.8397 | 0.5975 | 0.6163 | 0.6191 | 0.5438 | 0.6766 | 0.6627 | 0.6489 | 0.5970 | 0.5359 | 0.6316 | 0.6139 |
| MMreplacement [ 14 ] | 0.5622 | 0.5732 | 0.6284 | 0.6336 | 0.5038 | 0.4873 | 0.5467 | 0.5026 | 0.5384 | 0.4746 | 0.4908 | 0.5493 | 0.4624 | 0.5562 | 0.4672 |
| Prompt ID | Textual Description Prompt |
| 1 | The individual has shoulder-length brown hair styled with soft waves. They wear dangling earrings and a necklace, suggesting a formal or semi-formal occasion. Their expression is serious, with slightly parted lips as if mid-speech, indicating engagement in conversation. The posture appears upright, though only the head and shoulders are visible. The background is softly lit, featuring neutral tones that suggest an indoor setting, possibly a studio or event space. The overall appearance conveys professionalism and poise. |
| 2 | The individual has short, tousled brown hair and a stubbled beard, giving him a rugged appearance. His eyes are partially closed, and he has a slight smile, suggesting a relaxed or content mood. He is wearing a dark, casual shirt. The background is dimly lit, with indistinct shapes and colors, possibly indicating an evening event or gathering. His posture appears casual and confident. |
| 3 | The individual has short, graying hair with a receding hairline, a light beard, and a relaxed smile. His eyes are partially squinted, suggesting a friendly demeanor. He wears a black shirt and stands upright with a casual posture. The background features a blurred image of a desert landscape, possibly from a movie poster, indicating the setting might be a film-related event. The overall atmosphere is informal and approachable. |
| 4 | The individual has short, dark brown hair with a slight wave, styled neatly. Their face is oval-shaped with a strong jawline, and they have a calm, neutral expression. They are wearing a dark blue sweater over a white shirt. The posture is relaxed yet confident, leaning slightly forward. The background is blurred, suggesting an indoor setting with warm lighting, possibly a studio or home environment. The overall mood is serene and composed. |
| 5 | The individual wears a vibrant red and green patterned headwrap, complementing his blue shirt with a floral design. His face is adorned with a warm smile, showcasing a friendly demeanor. He has a robust build and appears to be in mid-conversation, suggesting an engaging personality. The blurred background hints at a public or social setting, possibly an event or gathering. His posture is upright, exuding confidence and approachability. The overall image conveys a sense of cultural pride and community involvement. |
| 6 | The individual has light skin, brown hair tied back, and is smiling widely, showing teeth, with eyes slightly squinted in joy. Their posture suggests movement, possibly dancing or celebrating, as one hand is raised near the face. The background is a bright blue sky, indicating an outdoor setting during daytime. The overall impression is one of happiness and spontaneity, captured in a candid moment. |
Deepfake Detection
Deepfake detection identifies manipulated or AI-generated media to support trustworthy communication, identity verification, and forensic analysis.
See all 264 papers