The quality and realism of synthetically generated fingerprint images have increased significantly over the past decade fueled by advancements in generative artificial intelligence (GenAI). This has exacerbated the vulnerability of fingerprint recognition systems to data injection attacks, where synthetic fingerprints are maliciously inserted during enrollment or authentication. Hence, there is an urgent need for methods to detect if a fingerprint image is real or synthetic. While it is straightforward to train deep neural network (DNN) models to classify images as real or synthetic, often such DNN models overfit the training data and fail to generalize well when applied to synthetic fingerprints generated using unseen GenAI models. In this work, we formulate synthetic fingerprint detection as a continual few-shot adaptation problem, where the objective is to rapidly evolve a base detector to identify new types of synthetic data. To enable continual few-shot adaptation, we employ a combination of binary cross-entropy and supervised contrastive (applied to the feature representation) losses and replay a few samples from previously known styles during fine-tuning to mitigate catastrophic forgetting. Experiments based on several DNN backbones (as feature extractors) and a variety of real and synthetic fingerprint datasets indicate that the proposed approach achieves a good trade-off between fast adaptation for detecting unseen synthetic styles and forgetting of known styles.
Figures & tables
Fig. 1: Vulnerability of fingerprint recognition systems (FRS) to data injection attacks. With advancements in GenAI, it is now possible to generate synthetic fingerprint images with any target style, which can be used to fool any FRS (especially in remote authentication scenarios) via bypassing of the sensor.
Fig. 2: Illustration of the proposed approach for synthetic fingerprint detection.
Dataset
Ink on Paper
Live Scan
Finger
Image
Finger
Image
Real Fingerprint Datasets
FVC-2000,02,04 [ 6 ]
-
-
330
7,920
NIST-SD4
2,000
4,000
-
-
NIST-SD14
27,000
54,000
-
-
NIST-SD300
17,659
17,659
-
-
TABLE I: Summary of fingerprint datasets used in this study. Here, “Finger” and “Image” columns denote the number of unique fingers and the total number of images, respectively.
Method
Adaptation Sequence
GenPrint
SFinGe
IBG-Novetta
IWGAN
PrintsGAN
FPGAN
Mean
TDR
FDR
TDR@
TDR
FDR
TDR@
TDR
FDR
TDR@
TDR
FDR
TDR@
TDR
FDR
TDR@
TDR
FDR
TDR@
TDR
FDR
TDR@
τ =0.5
τ =0.5
0.1% FDR
τ =0.5
τ =0.5
0.1% FDR
τ =0.5
τ =0.5
0.1% FDR
τ =0.5
τ =0.5
0.1% FDR
τ =0.5
τ =0.5
0.1% FDR
τ =0.5
τ =0.5
0.1% FDR
τ =0.5
τ =0.5
0.1% FDR
ViT-CE
GenPrint (base)
99.82
0.65
96.68
0.73
0.49
0.00
2.94
0.57
0.62
1.56
0.56
0.00
0.60
0.57
0.00
0.62
0.52
0.09
17.71
0.56
16.23
SFinGe
99.15
1.14
88.74
99.66
1.20
90.17
70.33
1.16
32.02
2.09
0.91
0.00
0.67
1.15
0.00
2.78
0.91
0.17
45.78
1.08
35.18
IBG-Novetta
98.56
1.66
82.06
99.91
1.91
91.54
99.06
1.66
80.10
1.64
1.67
0.00
2.19
1.89
0.09
1.92
1.81
0.02
50.55
1.77
42.30
IWGAN
99.28
0.85
91.16
99.87
0.85
98.46
99.21
0.85
94.97
94.17
0.89
44.77
20.30
0.92
3.05
2.81
1.11
0.17
69.27
0.91
55.43
TABLE II: Summary of the results from the continuous few-shot adaptation study. Here, each row represents an evolution of the detector - starting with the base detector followed by sequential adaptation to new synthetic styles. The main columns represents the performance on the test set of a known/unseen synthetic style, with the real test set remaining unchanged. The last column represents average performance over all the synthetic test sets.
Fig. 3: Impact of replay and sample size on continual few-shot adaptation of the base detector.
Fig. 4: Generalization to other DNN-backbones and sensitivity to weight ( λ ) assigned to the supervised contrastive (SC) loss.
Appendix figures & tables3 assets
Supplementary material from the paper’s appendix.
Appendix
Fig. 5: ROC curves (in log scale) illustrating the performance of continual few-shot adaptation. Each row corresponds to a different test set, while each column represents the incremental addition of data samples from each distribution. The orange boxes indicate the data distribution currently being added for adaptation. The black boxes represent data distributions that have already been incorporated and are under fine-tuning, while the gray boxes denote unseen data distributions. The magenta lines indicate the detector trained with CE+SupCon losses, while the blue lines represent the baseline trained using only the CE loss.
Fig. 6: Plots illustrating the evolution of feature representations for synthetic vs. real data. Each plot shows the corresponding synthetic samples along with all real data distributions, highlighting the effect of each adaptation step. The row and column layout matches the previous ROC plot. Each plot contains three subplots: (a) features color-coded based on ground-truth labels, (b) features color-coded based on predicted labels, and (c) features color-coded based on dataset name. In (a) and (b) , red denotes synthetic data features, while blue denotes real data features. In (c) colors correspond to datasets. Visualization suggests that few-shot adaptation primarily helps in forming the classifier boundary, while the underlying representation and separability remain largely unchanged and are mainly determined by the base detector.
Fig. 7: Figure illustrating the effect of dataset order in the adaptation sequence. Results are shown after the final adaptation step in continual training. In the left panel, each x-axis tick corresponds to a dataset, and each line represents a different adaptation sequence. The right panel organizes results with the actual adaptation sequence along the x-axis. Some variation in final performance is observed; however, it is not substantial, indicating that adaptation order has limited impact on final accuracy. In this few-shot setting, performance is primarily governed by the capacity of the base detector.
The rapid advancement of generative artificial intelligence (AI) has made synthetic images remarkably realistic, posing security threats such as misinformation and fraud. It is significant to detect the synthetic image in the manner of passive and blind image authentication. Most existing detectors rely on supervised training with large labeled datasets, leading to high costs and degraded performance on unknown generative models. To attenuate such deficiencies, we propose a training-free detection method. Specifically, noise residual fingerprints are first extracted by a simple yet effective pre-trained Noiseprint++ model. Then multi-scale features are further extracted from such residual by a frozen Vision Transformer (ViT), followed by adaptive weighted fusion. Only a few real image samples are used needed to initialize the clustering centers for unsupervised K-Means, distinguishing real and synthetic images without training. Extensive evaluations on four benchmark datasets show that our proposed scheme achieves an average accuracy of 82.2%, outperforming the state-of-the-art detectors on generalization ability. Superior performance is gained on the popular diffusion type of synthetic images, and the effectiveness of each module is validated by ablation studies. Source code will be publicly available at https://github.com/multimediaFor/NoiseCluSID.
Caihui Yan, Gang Cao, Huawei Tian +2
School of Computer and Cyber Sciences, Communication University of China, Beijing 100024, China · People's Public Security University of China, Beijing 100038, China
Contactless fingerprinting is an emerging approach to biometric authentication that allows users to scan their fingerprints without touching a scanner. Due to the limited amount of contactless fingerprint data available and the security risks associated with sharing real individuals' fingerprints, it is valuable to explore methods of generating synthetic data that can be used in place of - or in conjunction with - real data to develop and evaluate contactless fingerprinting systems. In this paper, we present and evaluate synthetic fingerphotos generated using StyleGAN2-ADA and StyleGAN3, existing image generation architectures. We evaluate the realism, privacy preservation, and variety of the synthetic fingerphotos by comparing their biometric feature statistics to those of real fingerphotos, computing match scores between real and synthetic fingerphotos, and computing match scores between different synthetic fingerphotos. This paper provides a quantitative comparison point for future evaluations of synthetic fingerphotos. The evaluation code is made available at https://github.com/cmillerlynch/fingerphoto-gan.
Conor Miller-Lynch, Sandip Purnapatra, Syed Konain Abbas +4
UNC Charlotte, Charlotte, NC, USA · Clarkson University, Potsdam, NY, USA
AI-generated image (AIGI) attribution presents a pressing challenge that goes beyond mere AIGI detection, aiming to identify the source model or technique responsible for a synthetic image. However, most previous source attribution methods operate in a closed-set manner, which necessitates retraining to recognize any novel category, preventing adaptation to the rapid evolution of image generation. In this work, we propose a new paradigm for synthetic image attribution, termed few-shot attribution. This paradigm targets the reliable identification of unseen generators using only limited samples, making it highly suitable for real-world applications. To facilitate this work, we construct OmniFake, a large-scale, well-categorized synthetic image dataset that contains 1.17 million images from 45 distinct generators. We further introduce OmniDFA (Omni Detector and Few-shot Attributor), a few-shot attribution baseline that not only assesses the authenticity of images but also determines their synthesis origins. Experiments demonstrate that OmniDFA exhibits excellent capability in few-shot attribution and achieves state-of-the-art generalization performance in AIGI detection. Our dataset and code are available at https://github.com/teheperinko541/OmniDFA.
Shiyu Wu, Shuyan Li, Jing Li +2
Institute of Automation, Chinese Academy of Sciences, Beijing, China · Beijing Academy of Artificial Intelligence, Beijing, China · University of Chinese Academy of Sciences, Beijing, China +3