Multimodal anti-spoofing in biometric recognition systems

Abstract

While multimodal biometric systems were commonly believed to be intrinsically more robust to spoof attacks than unimodal systems, recent results provided clear evidence that they can be evaded by spoofing a single biometric trait. This pointed out that also multimodal systems require specific anti-spoofing measures. In this chapter, we introduce the issue of multimodal anti-spoofing, and give an overview of state-of-the-art anti-spoofing measures. Such measures mainly consist of developing ad hoc score fusion rules that are based on assumptions about the match score distribution produced by fake biometric traits. We discuss the pros and cons of existing measures, and point out the current challenges in multimodal anti-spoofing.