指紋被廣泛運用在識別和驗證系統中，若系統被非法存取會造成很嚴重的安全問題。隨著生成對抗網路與擴散模型此生成式AI的發展，若被惡意使用會對系統安全性構成更重大的威脅。因此，開發具有泛化性的深度偽造指紋檢測方法並用於新的且訓練時未看過的偽造指紋變得極為重要。然而，現有的方法，包括通用檢測方法Universaldetection)和深度偽造檢測(deepfake) 方法，由於指紋圖片所能提供資訊有限而常常難以處理。這個挑戰在來自智慧手機側邊感測器和其他AIoT或邊緣設備的局部指紋上更加嚴峻，因為局部指紋所含之手指區域和像素大小都被更加縮減。此外，一些方法只能在多重偽造領域數據集上進行訓練，因資料收集問題使得實施更加複雜。相反地，儘管有些方法在單一領域訓練中表現良好，但在多領域訓練中表現不佳，使其不適合跨領域的泛化應用。
在這研究中，我們提出了MDMP-detector，是一種用於局部指紋圖像的泛化深度偽造檢測方法，其能支援並驗證在單一領域訓練和多領域訓練之中。我們引入了四個關鍵模組來實現。首先，為了改善偽造特徵域與真實特徵域之間的分離與距離的增加並增強特徵的泛化能力，我們採用了混合數據訓練(MixDataTraining)和內容相關信息去除Content-relative Information Removal) 模組。前者結合了局部和全局視角以強化資訊，後者則移除無用的內容相關資訊。為了進一步提升跨多域的泛化能力，提出了多重平行多分類器與特徵增強(MultipleParallelMulti-Classifier with Feature Augmentation)，其使用跨域拉和推增強(cross-domain pull and push augmentation) 來擴展偽造特徵分佈，增加未見偽造域其包含在此分佈下的可能性。隨著訓練偽造域數量的增加，跨域的泛化能力也會隨之增加。最後，引入了特徵增強分類器(MultipleParallelMulti-Classifier with Feature Augmentation)，以優化分類器的決策邊界，進一步來實現更好的泛化能力。
在我們的單域訓練評估結果中，我們的模型在真實數據與一個偽造數據集結合的訓練資料集上進行訓練，並在九個偽造數據集（其中五個分別來自不同的生成對抗網路模型，另外四個則來自擴散模型）上進行評估，達到88.15%的平均準確率(accuracy)，比UniversalFakeDetect [1] 高出 6.72%，且速度快了17.82倍。在多領域訓練中，使用兩個偽造數據集做訓練，準確率達到91.11%，比UCF[2]高出4.30%，速度快了12.37倍。使用三個偽造數據集進行訓練後，準確率進一步提升至92.63%，推理時間僅為每張圖像0.0161 秒。

Fingerprints are widely used for recognition and authentication, but unauthorized access poses serious security risks. The rapid evolution of generative AI, including GANs and diffusion models, intensifies this threat when misused. Therefore, developing generalized fingerprint deep forgery detection methods for unseen forgery domains is essential. However, existing methods, including universal and deepfake detection approaches, often struggle with fingerprints due to their limited information. This challenge is even greater with partial fingerprints from side mounted sensors on smartphones and other AIoT or edge devices, where both the finger area and pixel size are reduced. Additionally, some methods can only be trained on multiple forgery domain datasets, complicating implementation due to data collection issues. Conversely, while some methods perform well in single-domain training, they fail in multi-domain settings, making them unsuitable for cross-domain generalization.
In this work, we propose MDMP-detector,a generalized fingerprint deep forgery detection for partial fingerprints that supports both single-domain and multi-domain training on forgeries generated by GANs and diffusion models. We introduce four key modules to achieve this. To improve the separation and increase the distinction between the fake feature domains and the real feature domain and feature generalization, we employ Mix Data Training and Content-Related Information Removal. The former integrates local and global views to enhance information, and the letter remove useless content information. To further boost generalization across multiple domains, we propose Multiple Parallel Multi-Classifier with Feature Augmentation. Using cross-domain pull and push augmentation to broaden forgery feature distributions, enhancing the likelihood of including unseen forgery domains. The generalization ability across multiple domains improves as more forgery domains are included in training. Finally, a Classifier with Feature Augmentation is introduced to refine the decision boundary of the classifier, enhancing better generalization.
In single-domain training, our model, trained on a real dataset combined with one generated forgery dataset and evaluated on nine generated forgery datasets(five GANs and four diffusions), achieved an average accuracy of 88.15%, 6.72% higher than UniversalFakeDetect [1] and 17.82 times faster. In multiple-domain training with two forgery datasets, it reached 91.11% accuracy, outperforming UCF [2] by 4.30% and running 12.37 times faster. Training on three forgery datasets further improved accuracy to 92.63%, with an inference time of just 0.0161 seconds per image.

[1] U. Ojha, Y. Li, and Y. J. Lee, “Towards universal fake image detectors that generalize
across generative models,” in Proceedings of the IEEE/CVF Conference on Computer
Vision and Pattern Recognition, pp. 24480–24489, 2023.
[2] Z. Yan, Y. Zhang, Y. Fan, and B. Wu, “Ucf: Uncovering common features for generaliz
able deepfake detection,” in Proceedings of the IEEE/CVF International Conference on
Computer Vision, pp. 22412–22423, 2023.
[3] M.-H. Hsu, Y.-C. Hsu, and C.-T. Chiu, “Tiny partial fingerprint sensor quality assess
ment,” IEEE Sensors Letters, 2024.
[4] A.Rattani, W. J. Scheirer, and A. Ross, “Open set fingerprint spoof detection across novel
fabrication materials,” IEEE Transactions on Information Forensics and Security, vol. 10,
no. 11, pp. 2447–2460, 2015.
[5] Y. Ding and A. Ross, “An ensemble of one-class svms for fingerprint spoof detection
across different fabrication materials,” in 2016 IEEE International Workshop on Informa
tion Forensics and Security (WIFS), pp. 1–6, IEEE, 2016.
[6] C. Yuan, Z. Xia, X. Sun, and Q. J. Wu, “Deep residual network with adaptive learning
framework for fingerprint liveness detection,” IEEE Transactions on Cognitive and De
velopmental Systems, vol. 12, no. 3, pp. 461–473, 2019.
[7] Z. Xia, C. Yuan, R. Lv, X. Sun, N. N. Xiong, and Y.-Q. Shi, “A novel weber local binary
descriptor for fingerprint liveness detection,” IEEE Transactions on Systems, Man, and
Cybernetics: Systems, vol. 50, no. 4, pp. 1526–1536, 2018.
[8] H. Liu, W. Zhang, F. Liu, H. Wu, and L. Shen, “Fingerprint presentation attack detector
using global-local model,” IEEE Transactions on Cybernetics, vol. 52, no. 11, pp. 12315
12328, 2021.
[9] S. A. Grosz and A. K. Jain, “Spoofgan: Synthetic fingerprint spoof images,” IEEE Trans
actions on Information Forensics and Security, vol. 18, pp. 730–743, 2022.
[10] A. S. J. GAtul K. Uttam, Rohit Agarwal, “Fingerprint presentation attack detection using
unsuperviseddivergencebaseddomainadaptation,” JournalofElectricalSystems, vol.20,
pp. 2522–2531, 2024.
[11] A. Shoshan, N. Bhonker, E. Ben Baruch, O. Nizan, I. Kviatkovsky, J. Engelsma, M. Ag
garwal, and G. Medioni, “Fpgan-control: A controllable fingerprint generator for training
with synthetic data,” in Proceedings of the IEEE/CVF Winter Conference on Applications
of Computer Vision, pp. 6067–6076, 2024.
[12] H. Miao, Y. Guo, and Y. Wang, “Rfdforfin: Robust deep forgery detection for gan
generated fingerprint images,” arXiv preprint arXiv:2308.09285, 2023.
[13] J.-Y. Zhu, T. Park, P. Isola, and A. A. Efros, “Unpaired image-to-image translation using
cycle-consistent adversarial networks,” in Proceedings of the IEEE international confer
ence on computer vision, pp. 2223–2232, 2017.
[14] C. Dong, A. Kumar, and E. Liu, “Think twice before detecting gan-generated fake im
ages from their spectral domain imprints,” in Proceedings of the IEEE/CVF conference
on computer vision and pattern recognition, pp. 7865–7874, 2022.
[15] D. Maio, D. Maltoni, R. Cappelli, J. L. Wayman, and A. K. Jain, “Fvc2004: Third finger
print verification competition,” pp. 1–7, 2004.
[16] R. Cappelli, M. Ferrara, A. Franco, and D. Maltoni, “Fingerprint verification competition
2006,” Biometric Technology Today, vol. 15, no. 7-8, pp. 7–9, 2007.
[17] G. P. Fiumara, P. A. Flanagan, J. D. Grantham, K. Ko, K. Marshall, M. Schwarz,
E. Tabassi, B. Woodgate, and C. Boehnen, “Nist special database 302: Nail to nail fin
gerprint challenge,” 2019.
[18] G. Fiumara, P. Flanagan, M. Schwarz, E. Tabassi, and C. Boehnen, “Nist special database
301,” Gaithersburg, MD, USA, 2018.
[19] C. Tan, Y. Zhao, S. Wei, G. Gu, P. Liu, and Y. Wei, “Frequency-aware deepfake detection:
Improving generalizability through frequency space domain learning,” in Proceedings of
the AAAI Conference on Artificial Intelligence, vol. 38, pp. 5052–5060, 2024.
[20] C. Tan, Y. Zhao, S. Wei, G. Gu, and Y. Wei, “Learning on gradients: Generalized arti
facts representation for gan-generated images detection,” in Proceedings of the IEEE/CVF
Conference on Computer Vision and Pattern Recognition, pp. 12105–12114, 2023.
[21] C.Tan,Y.Zhao,S.Wei,G.Gu,P.Liu,andY.Wei,“Rethinkingtheup-samplingoperations
in cnn-based generative network for generalizable deepfake detection,” in Proceedings
of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 28130
28139, 2024.
[22] Y. Jia, J. Zhang, S. Shan, and X. Chen, “Single-side domain generalization for face anti
spoofing,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern
Recognition, pp. 8484–8493, 2020.
[23] T. Karras, T. Aila, S. Laine, and J. Lehtinen, “Progressive growing of gans for improved
quality, stability, and variation,” arXiv preprint arXiv:1710.10196, 2017.
[24] T. Karras, S. Laine, and T. Aila, “A style-based generator architecture for generative ad
versarial networks,” in Proceedings of the IEEE/CVF conference on computer vision and
pattern recognition, pp. 4401–4410, 2019.
[25] Q. Chen and V. Koltun, “Photographic image synthesis with cascaded refinement
networks,” in Proceedings of the IEEE international conference on computer vision,
pp. 1511–1520, 2017.
[26] A. Rossler, D. Cozzolino, L. Verdoliva, C. Riess, J. Thies, and M. Nießner, “Faceforen
sics++: Learning to detect manipulated facial images,” in Proceedings of the IEEE/CVF
international conference on computer vision, pp. 1–11, 2019.
[27] A. Nichol, P. Dhariwal, A. Ramesh, P. Shyam, P. Mishkin, B. McGrew, I. Sutskever, and
M. Chen, “Glide: Towards photorealistic image generation and editing with text-guided
diffusion models,” arXiv preprint arXiv:2112.10741, 2021.
[28] A. Ramesh, M. Pavlov, G. Goh, S. Gray, C. Voss, A. Radford, M. Chen, and I. Sutskever,
“Zero-shot text-to-image generation,” in International conference on machine learning,
pp. 8821–8831, Pmlr, 2021.
[29] S.-Y. Wang, O. Wang, R. Zhang, A. Owens, and A. A. Efros, “Cnn-generated images
are surprisingly easy to spot... for now,” in Proceedings of the IEEE/CVF conference on
computer vision and pattern recognition, pp. 8695–8704, 2020.
[30] A. Radford, J. W. Kim, C. Hallacy, A. Ramesh, G. Goh, S. Agarwal, G. Sastry, A. Askell,
P. Mishkin, J. Clark, et al., “Learning transferable visual models from natural language
supervision,” in International conference on machine learning, pp. 8748–8763, PMLR,
2021.
[31] Y. Jeong, D. Kim, S. Min, S. Joe, Y. Gwon, and J. Choi, “Bihpf: Bilateral high-pass filters
for robust deepfake detection,” in Proceedings of the IEEE/CVF Winter Conference on
Applications of Computer Vision, pp. 48–57, 2022.
[32] Y. Jeong, D. Kim, Y. Ro, and J. Choi, “Frepgan: robust deepfake detection using
frequency-level perturbations,” in Proceedings of the AAAI conference on artificial in
telligence, vol. 36, pp. 1060–1068, 2022.
[33] Z. Liu, H. Wang, and S. Wang, “Cross-domain local characteristic enhanced deepfake
video detection,” in Proceedings of the Asian Conference on Computer Vision, pp. 3412
3429, 2022.
[34] J. Cao, C. Ma, T. Yao, S. Chen, S. Ding, and X. Yang, “End-to-end reconstruction
classification learning for face forgery detection,” in Proceedings of the IEEE/CVF Con
ference on Computer Vision and Pattern Recognition, pp. 4113–4122, 2022.
[35] Z. Yan, Y. Luo, S. Lyu, Q. Liu, and B. Wu, “Transcending forgery specificity with la
tent space augmentation for generalizable deepfake detection,” in Proceedings of the
IEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 8984–8994,
2024.
[36] T. Yang, J. Cao, Q. Sheng, L. Li, J. Ji, X. Li, and S. Tang, “Learning to disentangle gan
fingerprint for fake image attribution,” arXiv preprint arXiv:2106.08749, 2021.
[37] Hsu0623, “Inpainting-diffusion-synthetic-and-data-augment-with-feature-keypoints-for
tiny-partial-fingerprints,” 2024.
[38] Y. Lu, Y.-W. Tai, and C.-K. Tang, “Attribute-guided face generation using conditional cy
clegan,” in Proceedings of the European conference on computer vision (ECCV), pp. 282
297, 2018.
[39] T. Karras, M. Aittala, J. Hellsten, S. Laine, J. Lehtinen, and T. Aila, “Training genera
tive adversarial networks with limited data,” Advances in neural information processing
systems, vol. 33, pp. 12104–12114, 2020.
[40] J. Song, C. Meng, and S. Ermon, “Denoising diffusion implicit models,” arXiv preprint
arXiv:2010.02502, 2020.
[41] P. Dhariwal and A. Nichol, “Diffusion models beat gans on image synthesis,” Advances
in neural information processing systems, vol. 34, pp. 8780–8794, 2021.
[42] R. Rombach, A. Blattmann, D. Lorenz, P. Esser, and B. Ommer, “High-resolution image
synthesis with latent diffusion models,” in Proceedings of the IEEE/CVF conference on
computer vision and pattern recognition, pp. 10684–10695, 2022.
[43] V. Verma, A. Lamb, C. Beckham, A. Najafi, I. Mitliagkas, D. Lopez-Paz, and Y. Bengio,
“Manifold mixup: Better representations by interpolating hidden states,” in International
conference on machine learning, pp. 6438–6447, PMLR, 2019.