Abstract

Facial recognition is a popular biometric security system used to authenticate individuals based on their unique facial structure. However, this system is vulnerable to spoofing attacks where the attacker can bypass the system using fake representations of the user's face such as photos, statues or videos. Liveness detection is a method used to address this issue by verifying that the user is a real person and not a representation. This journal article focuses on the life sign method of liveness detection, which utilizes facial movements to confirm the user's existence. We implement the latest technology of artificial intelligence from TensorFlow.js using face-api.js and compare it with the GLCM algorithm. However, even with the life sign detection method, there is still a chance of bypassing the system if an attacker uses a video recording. To mitigate this, we propose the addition of an object detection system to detect the hardware used to show video recordings with ml5.js. Our face recognition and expression detection system, using the pre-trained model face-api.js, achieved an accuracy of 85% and 82.5%, respectively, and the object detection system built with ml5.js has high accuracy and is very effective for liveness detection. Our results indicate that face-api.js outperformed GLCM algorithm in detecting spoofing attempts.

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call