Abstract:
From improving security procedures in sensitive areas to unlocking personal devices, facial recognition systems are now essential to many applications. Their non-intrusive authentication process and user-friendliness are the reasons behind their widespread adoption. The integrity of these systems is seriously threatened by the increase in spoofing attacks, in which adversaries use fake facial representations like images, videos, or masks. This thesis offers a thorough investigation into creating a strong face anti-spoofing system with deep learning methods. To improve decision-making accuracy, the suggested method combines ensemble learning techniques, support vector machines (SVMs) for classifcation, and convolutional neural networks (CNNs) for feature extraction. Because CNNs are skilled at identifying complex patterns in facial data, discriminative features that differentiate between real and fake faces can be extracted. To correctly classify these features, support vector machines (SVMs), which are well-known for their performance in binary classifcation tasks, are used. Ensemble learning techniques, which combine the strengths of multiple classifers to improve overall performance, are incorporated to further strengthen the system’s resilience against various spoofng attacks. The effectiveness of the suggested system is confrmed by extensive testing on benchmark datasets, such as Replay-Attack and CASIA-FASD. These fndings show that the combination of CNNs, SVMs, and ensemble learning techniques greatly improves the system’s detection and prevention of spoofng attempts in a variety of attack scenarios. This demonstrates how the suggested method could strengthen the security of facial recognition apps and guarantee safe and dependable authentication in practical applications.
