Apple in my eyes (AIME): liveness detection for mobile security using corneal specular reflections

ABSTRACTIn this paper, we present a novel software-based face Presentation Attack Detection (PAD) method named "Apple in My Eyes (AIME)" using screen display as a challenge and corneal specular reflections as a response for authenticating the liveness against presentation. To detect face liveness, AIME creates multiple image patterns on the authentication screen as a challenge, then captures meaningful corneal specular reflection responses from user's eyes using the front camera, and analyzes the reflective pattern images using various lightweight Machine Learning (ML) techniques under a subsecond level delay (200 ms). We demonstrate that AIME can detect various attacks, including digital images displayed on the phone or tablet, printed paper images, 2D paper masks, videos, 3D silicon masks, and 3D facial models using VR. AIME liveness detection can be applied for various contactless biometric authentication accurately and efficiently without any costly extra sensors.