Face recognition is used in biometric systems to verify and authenticate an individual. However, most face authentication systems are prone to spoofing attacks such as replay attacks, attacks using 3D masks etc. Thus, the importance of face anti-spoofing algorithms is becoming essential in these systems. Recently, deep learning has emerged and achieved excellent results in challenging tasks related to computer vision. The proposed framework relies on the extraction of features from the faces of individuals. The approach relies on dimensionality reduction and feature extraction of input frames using pre-trained weights of convolutional autoencoders, followed by classification using softmax classifier. Experimental analysis on three benchmarks, Idiap Replay Attack, CASIA- FASD and 3DMAD, shows that the proposed framework can attain results comparable to state-of-the-art methods in both cross-database and intra-database testing.

人脸识别在生物识别系统中用于验证和认证个人。但是，大多数人脸认证系统都容易遭受欺骗攻击，例如重播攻击，使用3D掩码的攻击等。因此，在这些系统中，人脸反欺骗算法的重要性变得至关重要。最近，在与计算机视觉有关的艰巨任务中，深度学习已经出现并取得了优异的成绩。所提出的框架依赖于从个人面孔中提取特征。该方法依赖于使用卷积自动编码器的预训练权重的降维和输入帧特征提取，然后使用softmax分类器进行分类。对三个基准测试（Idiap重播攻击，CASIA-FASD和3DMAD）进行实验分析，