Anti-Spoofing Techniques in Face Recognition, an Ensemble Based Approach

摘要

In this article we describe the implementation of a reliable and innovative ensemble-based technique that can prevent face spoofing attacks. The presented software is part of a technology developed in partnership with IsItYou, an Israeli company, that attempts to replace passwords with a face-based authentication system. Since the main problem of biometric systems is represented by the spoof attacks, IsItYou came with a solution to this, developing a unique technology that can identify spoof attacks, and authenticate only authorized humans. Inspired from deep learning techniques where ensemble-based solutions improve machine learning results by uniting several models, a software ensemble that combines multiple anti-spoofing methods, covering a larger range of spoof attacks and increasing overall security was developed. The article also shows the performances results and implementation details. The experimental results signpost our solution can provide first-rate results compared to the state-of-the-art approaches.