Multi-Modal Face Anti-Spoofing Identification Using Efficient Networks Integrated with U-Adapter for Unreliable Region Mitigation
| International Journal of Electronics and Communication Engineering |
| © 2025 by SSRG - IJECE Journal |
| Volume 12 Issue 11 |
| Year of Publication : 2025 |
| Authors : Mudunuru Suneel, Tummala Ranga Babu |
10.14445/23488549/IJECE-V12I11P111
How to Cite?
Mudunuru Suneel, Tummala Ranga Babu, "Multi-Modal Face Anti-Spoofing Identification Using Efficient Networks Integrated with U-Adapter for Unreliable Region Mitigation," SSRG International Journal of Electronics and Communication Engineering, vol. 12, no. 11, pp. 133-145, 2025. Crossref, https://doi.org/10.14445/23488549/IJECE-V12I11P111
Abstract:
A face anti-spoofing system that identifies a live face from a fake face done via photographs, video playback clips, or 3D masks should strengthen secure facial recognition systems. This development presents a new, innovative, multi-modal face anti-spoofing system using EfficientNet as the backbone architecture, with input signals comprising RGB, depth, and infrared. The suggested approach utilizes three separate EfficientNet models working side by side in order to operate on visibly disparate features for each of the modes, which should then be tailored for anti-spoofing without introducing additional complexity into the network using U-Adapters, which are lightweight modules designed to do so. This is achieved through the introduction of a ReGrad block, which aims to balance the importance of features extracted by different modes by regulating their gradient contributions. Once adapted and extracted, the outputs from all these U-Adapters are concatenated, creating one common format that merges major details from RGB, depth, and infrared signals. Lastly, a final softmax layer is used to classify whether the face is genuine or spoofed. On benchmark datasets, it is shown in experiments that the model is effective as it hit 98.3% accuracy on the CelebA-Spoof dataset and has demonstrated more robustness against diverse spoofing techniques than single-modal or other multi-modal models. Ablation studies indicated that the key results were achieved by employing U-Adapters together with the ReGrad block, where all these components had an impact on improving precision and generalization, respectively. Consequently, it was reported that the role of rebalancing gradient significance across modes played by the ReGrad block was important enough to enhance its adaptivity against the influence of the environment. This model improves on facial fabbing accuracy beyond anything that has been achieved so far; on top of that, it shows why efficient multimodal processing is important. Further research may also try this method in areas without some modes of modal features or different modalities that might be considered to make it more robust against spoofing techniques.
Keywords:
Multi-modal face anti-spoofing, EfficientNet, U-Adapter, ReGrad.
References:
[CrossRef] [Google Scholar] [Publisher Link]
[2] N. Dalal, and B. Triggs, “Histograms of Oriented Gradients for Human Detection,” 2005 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR'05), San Diego, CA, USA, vol. 1, pp. 886-893, 2005.
[CrossRef] [Google Scholar] [Publisher Link]
[3] Jukka Komulainen, Abdenour Hadid, and Matti Pietikäinen, “Context based Face Anti-Spoofing,” 2013 IEEE Sixth International Conference on Biometrics: Theory, Applications and Systems (BTAS), Arlington, VA, USA, pp. 1-8, 2013.
[CrossRef] [Google Scholar] [Publisher Link]
[4] Omkar M. Parkhi, Andrea Vedaldi, and Andrew Zisserman, “Deep Face Recognition,” Procedings of the British Machine Vision Conference, pp. 1-12, 2015.
[CrossRef] [Google Scholar] [Publisher Link]
[5] Bowen Zhang, Benedetta Tondi, and Mauro Barni, “Adversarial Examples for Replay Attacks Against CNN-based Face Recognition with Anti-Spoofing Capability,” Computer Vision and Image Understanding, vol. 197-198, 2020.
[CrossRef] [Google Scholar] [Publisher Link]
[6] Jianwei Yang et al., “Face Liveness Detection with Component Dependent Descriptor,” 2013 International Conference on Biometrics (ICB), Madrid, Spain, pp. 1-6, 2013.
[CrossRef] [Google Scholar] [Publisher Link]
[7] Anjith George, and Sébastien Marcel, “Cross Modal Focal Loss for RGBD Face Anti-Spoofing,” 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Nashville, TN, USA, pp. 7878-7887, 2021.
[CrossRef] [Google Scholar] [Publisher Link]
[8] Jun Liu, and Ajay Kumar “Detecting Presentation Attacks from 3D Face Masks Under Multispectral Imaging,” 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW), Salt Lake City, UT, USA, pp. 47-475, 2018.
[CrossRef] [Google Scholar] [Publisher Link]
[9] Zitong Yu et al., “Multi-Modal Face Anti-Spoofing Based on Central Difference Networks,” 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW), Seattle, WA, USA, pp. 2766-2774, 2020.
[CrossRef] [Google Scholar] [Publisher Link]
[10] Zitong Yu et al., “Flexible-Modal Face Anti-Spoofing: A Benchmark,” 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW), Vancouver, BC, Canada, pp. 6346-6351, 2023.
[CrossRef] [Google Scholar] [Publisher Link]
[11] Mingxing Tan, and Quoc Le “EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks,” Proceedings of the 36th International Conference on Machine Learning, pp. 6105-6114, 2019.
[Google Scholar] [Publisher Link]
[12] Yuanhan Zhang et al., “CelebA-spoof: Large-Scale Face Anti-Spoofing Dataset with Rich Annotations,” Computer Vision–ECCV 2020: 16th European Conference, Glasgow, UK, pp. 70-85, 2020.
[CrossRef] [Google Scholar] [Publisher Link]
[13] Shifeng Zhang et al., “Casia-surf: A Large-Scale Multi-Modal Benchmark for Face Anti-Spoofing,” IEEE Transactions on Biometrics, Behavior, and Identity Science, vol. 2, no. 2, pp. 182-193, 2020.
[CrossRef] [Google Scholar] [Publisher Link]
[14] Anjith George et al., “Biometric Face Presentation Attack Detection with Multi-Channel Convolutional Neural Network,” IEEE Transactions on Information Forensics and Security, vol. 15, pp. 42-55, 2020.
[CrossRef] [Google Scholar] [Publisher Link]
[15] Di Wen, Hu Han, and Anil K. Jain, “Face Spoof Detection with Image Distortion Analysis,” IEEE Transactions on Information Forensics and Security, vol. 10, no. 4, pp. 746-761, 2015.
[CrossRef] [Google Scholar] [Publisher Link]
[16] Zitong Yu et al., “Searching Central Difference Convolutional Networks for Face Anti-Spoofing,” 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Seattle, WA, USA, pp. 5295-5305, 2020.
[CrossRef] [Google Scholar] [Publisher Link]
[17] Anjith George, and Sébastien Marcel, “Cross Modal Focal Loss for RGBD Face Antispoofing,” 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Nashville, TN, USA, pp. 7882-7891, 2021.
[CrossRef] [Google Scholar] [Publisher Link]
[18] Aleksandr Parkin, and Oleg Grinchuk, “Recognizing Multi-Modal Face Spoofing with Face Recognition Networks,” 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW), Long Beach, CA, USA, pp. 1617-1623, 2019.
[CrossRef] [Google Scholar] [Publisher Link]
[19] Zhuo Wang et al., “Face Anti-Spoofing using Transformers with Relation-Aware Mechanism,” IEEE Transactions on Biometrics, Behavior, and Identity Science, vol. 4, no. 3, pp. 439-450, 2022.
[CrossRef] [Google Scholar] [Publisher Link]
[20] Chelsea Finn, Pieter Abbeel, and Sergey Levine, “Model-Agnostic Meta-Learning for Fast Adaptation of Deep Networks,” Proceedings of the 34th International Conference on Machine Learning, Sydney NSW Australia, vol. 70, pp. 1126-1135, 2017.
[Google Scholar] [Publisher Link]
[21] Jukka Määttä, Abdenour Hadid, and Matti Pietikäinen, “Face Spoofing Detection from Single Images using Micro-Texture Analysis,” 2011 International Joint Conference on Biometrics (IJCB), Washington, DC, USA, pp. 1-7, 2011.
[CrossRef] [Google Scholar] [Publisher Link]
[22] Oeslle Lucena et al., “Transfer Learning using Convolutional Neural Nettworks for Face Anti-Spoofing,” International Conference on Image Analysis and Recognition, pp. 27-34, 2017.
[CrossRef] [Google Scholar] [Publisher Link]
[23] Zezheng Wang et al., “Deep Spatial Gradient and Temporal Depth Learning for Face Antispoofing,” 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), Seattle, WA, USA, pp. 5042-5050, 2020.
[CrossRef] [Google Scholar] [Publisher Link]
[24] Mudunuru Suneel, and Tummala Ranga Babu, “Efficient Face Anti-Spoofing Identifier Network (FASIN) with Depth and Near Infrared Deep Learning Methods,” ARPN Journal of Engineering and Applied Sciences, vol. 19, no. 20, pp. 1255-1265, 2024.
[Google Scholar] [Publisher Link]
[25] Mudunuru Suneel, and Tummala Ranga Babu, “Spoof-Formernet: The Face Anti Spoofing Identifier with a Two Stage High Resolution Vision Transformer (HR-ViT) Network,” International Journal of Image, Graphics and Signal Processing, vol. 17, no. 4, pp. 87-104, 2025.
[CrossRef] [Google Scholar] [Publisher Link]