Efficient software attack to multimodal biometric systems and its application to face and iris fusion

In certain applications based on multimodal interaction it may be crucial to determine not only what the user is doing (commands), but who is doing it, in order to prevent fraudulent use of the system. The biometric technology, and particularly the multimodal biometric systems, represent a highly efficient automatic recognition solution for this type of applications. Although multimodal biometric systems have been traditionally regarded as more secure than unimodal systems, their vulnerabilities to spoofing attacks have been recently shown. New fusion techniques have been proposed and their performance thoroughly analysed in an attempt to increase the robustness of multimodal systems to these spoofing attacks. However, the vulnerabilities of multimodal approaches to software-based attacks still remain unexplored. In this work we present the first software attack against multimodal biometric systems. Its performance is tested against a multimodal system based on face and iris, showing the vulnerabilities of the system to this new type of threat. Score quantization is afterwards studied as a possible countermeasure, managing to cancel the effects of the proposed attacking methodology under certain scenarios.