Article Details

Abstract

With the rapid development of artificial intelligence and digital image processing technology, maliciously tampered face videos spread widely on the Internet, causing non-negligible security hazards and ethical challenges to social life. Traditional face tampering detection methods rely on appearance texture and deep learning features, and are vulnerable to illumination interference, difficult to defend against high-fidelity deepfake and 3D mask attacks, and lacking in the mining of robust living-body features. To solve the above defects, this paper proposes a face tampering detection method based on facial pulse physiological features. Relying on the principle of remote photoplethysmography, the proposed method extracts weak temporal pulse signals from effective facial regions, converts time-domain signals into frequency-domain features via Fourier transform, and realizes face liveness detection using a support vector machine model. Compared with traditional texture feature extraction methods, the proposed method adopts unfalsifiable physiological signals as the discrimination criterion, which can effectively identify various face tampering attacks, including photo replay, screen shooting and 3D face forgery. Experimental results show that the proposed framework achieves an accuracy of 86.3%, a precision of 84.0%, a recall of 82.0%, and an F1-score of 83.0% on the 3DMAD dataset. The method outperforms existing related methods and can effectively resist attacks from printed photos, video replays and 3D masks.

Keywords

face tampering detection; remote photoplethysmography; frequency domain feature; liveness detection

References

1. [1] LIU Fei, ZHANG Kunbo, YANG Qing, et al. A Survey of 3D Face Imaging and Reconstruction Technology [J]. Journal of Image and Graphics, 2024, 29(9): 2441-2470.
2. [2] XU Zeli, YU Kuidong. Inspection of Copy-Move Forgery Images Based on LBP [J]. Forensic Science and Technology, 2023, 48(5): 481-488.
3. [3] Sabeena M, Abraham L, Varghese A. Digital Image Forgery Detection Using Local Binary Pattern (LBP) and Harlick Transform with Classification [C]//2021 IEEE International Power and Renewable Energy Conference (IPRECON), 2021: 1-6.
4. [4] Passos L A, Jodas D, Costa K A P, et al. A Review of Deep Learning-based Approaches for Deepfake Content Detection [J/OL]. Expert Systems, 2024, 41(8): e13570. https://doi.org/10.1111/exsy.13570.
5. [5] SUI Shenyang, WANG Yongxiong, GENG Mengqing. Remote Heart Rate Estimation Based on Frequency Regulation and Multi-Scale Spatio-Temporal Modeling [J]. Journal of Biomedical Engineering, 2026, 43(1): 138-145.
6. [6] Yang Z, Wang H F, Lu F. Assessment of Deep Learning-Based Heart Rate Estimation Using Remote Photoplethysmography Under Different Illuminations [J]. IEEE Transactions on Human-Machine Systems, 2022, 52(6): 1236-1246.
7. [7] TIAN Huijuan, ZHANG Yajing, QIAN Sheng, et al. Heart Rate Detection Method Based on Multi-Color-Space Illumination Invariance [J]. Journal of Tianjin Polytechnic University, 2025, 44(4): 28-35.
8. [8] Przybyło J. A Deep Learning Approach for Remote Heart Rate Estimation [J/OL]. Biomedical Signal Processing and Control, 2022, 74: 103457. https://doi.org/10.1016/j.bspc.2021.103457.
9. [9] Almarshad M A, Islam M S, Al-Ahmadi S, et al. Diagnostic Features and Potential Applications of PPG Signal in Healthcare: A Systematic Review [J/OL]. Healthcare, 2022, 10(3): 547. https://doi.org/10.3390/healthcare10030547.
10. [10] Nayak S K, Pradhan B, Mohanty B, et al. A Review of Methods and Applications for a Heart Rate Variability Analysis [J]. Algorithms, 2023, 16(9): 433. https://doi.org/10.3390/a16090433.
11. [11] Lundstrom C J, Foreman N A, Biltz G. Practices and Applications of Heart Rate Variability Monitoring in Endurance Athletes [J]. International Journal of Sports Medicine, 2023, 44(1): 9-19.
12. [12] Yao X, Song R C, Cheng J, et al. HRVFusion: Video-Based Long-Term Heart Rate Variability Measurement With Conditional Diffusion Models [J]. IEEE Transactions on Multimedia, 2026, 28: 7371-7384.
13. [13] De Haan G, Jeanne V. Robust Pulse Rate From Chrominance-Based rPPG [J]. IEEE Transactions on Biomedical Engineering, 2013, 60(10): 2878-2886.

Funding

Guangdong Medical Science and Technology Research Foundation Project of Guangdong Provincial Health Commission (B2026258); 2025 Guangdong Vocational Digital Transformation Steering Committee Teaching Reform and Research Project (2025SZXZH007); Guangdong Food and Drug Vocational College Project on Building a New Digital Teaching Ecosystem (2025JGZX04, 2025JGZX07); Guangdong Food and Drug Vocational College Education and Teaching Reform Project (2025JGZX02); Guangdong Medical Science and Technology Research Foundation Project of Guangdong Provincial Health Commission (A2024402); Guangdong Provincial Education Science Planning Project (2024GXJK182); Guangdong Food and Drug Vocational College Natural Science Research Project (2024ZR13)

About the authors

BIAN Mingyun (1993—), male, Han ethnicity, born in Jining, Shandong, lecturer, master's supervisor, doctor. Research interests: privacy preservation and AI security.