DEVELOPMENT OF PASSIVE LIVENESS DETECTION SYSTEM BASED ON DEEP LEARNING LIVENESSNET TO OVERCOME FACE SPOOFING

• Main Author: Shafa Alya, Dira
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/74190
• Institution: Institut Teknologi Bandung

In an effort to increase Indonesia's competitiveness in the digital era, the government through the Sistem Pemerintahan Berbasis Elektronik (SPBE) program has launched a digital government system. With the implementation of SPBE, a more efficient and transparent governance system is expected to be achieved. However, there are challenges such as identity and personal data risks to overcome. By using a biometric-based identity verification system such as face recognition, the security risk can be minimized. On the other hand, the distribution of personal information on various internet sites is increasing. This triggers many incidents of identity theft by misusing information spread on the internet. One example of identity attack is face spoofing, face spoofing is an attempt by someone to falsify the face recognition system in order to gain access to the system in the wrong way. Therefore, we developed a Passive Liveness Detection system based on deep learning model, LivenessNet, which can detect real and spoofing faces as additional security to the facial recognition system. Types of spoofing that can be overcome with a Passive Liveness Detection system are 2D-based like print attacks and replay attacks. Based on the tests conducted to LivenessNet model, the final model achieved an accuracy of 97.17%, False Positive Rate (FPR) of 0%, and False Negative Rate (FNR) of 5.68% on a custom dataset containing 3000 images. For larger datasets, 90% accuracy, 7.16% False Positive Rate (FPR), and 28.48% False Negative Rate (FNR) are obtained. Thus, the resulting LivenessNet model can meet the specified system specifications.