Multibiometric system based on Face Video and Heart Rate Estimation

Biometric systems have acquired a central role in identity recognition because of the security and convenience they offer over traditional authentication systems such as those based on the knowledge of passwords or the possession of tokens. Multibiometric systems have been introduced to increase the performance of mono-modal systems, to increase their security and to make the systems more universal. The diffusion of biometric systems has stimulated the attempts to fraudulent access. Adversarial Examples (AE) belong to the most effective category of attacks. They are input samples specifically designed to fool a target biometric system: they can be implemented using several approaches and algorithms. Therefore, appropriate and effective defense strategies need to be developed. Despite the researchers carried out different solutions, it is still an open challenge. Recently, researchers described how it is possible to extract several physiological parameters from videos containing faces. In fact, imperceptible movements for humans contain information on breathing rate and emotional status, while the blood flow produces an effect on the light diffused from the face skin. The latter can be acquired using RGB camera to extract Hearth Rate (HR) and the Photoplethysmogram (PPG) signal. This research is twofold: the development a multibiometric system based on face video and HR estimation to assess the performance obtained both in verification and identification task and the evaluation of robustness against AE.