The aims of the project is to develop reliable and efficient solutions to detect and circumvent photo replay attacks in facial authentication systems. Typical countermeasures to spoofing attacks, including the use of dedicated sensors, challenge-response methods or multiple biometric modality combination cannot be deployed in scenarios where the only available sensor is the user camera. The new solution will then only rely on the video stream signal to detect spoofing attacks.
The typical countermeasure to a spoofing attack is liveness detection that aims at detecting physiological signs of life. This can be done in different ways, for instance with dedicated sensors (ex: multi-spectral imagery) to detect an evidence of liveness, with a challenge response method where a spoofing attack can be detected by asking the user to interact with the system (ex: text prompted speech recognition) or with multi-modal biometrics (ex: joint face and voice recognition). None of these typical countermeasures can be considered in the framework because the only available sensor is the camera of the user. The innovation here consists then in the development of reliable anti-spoofing solutions, based on the detection of characteristic pattern of liveness/spoofing on the only available signal: the video stream of the camera.
The results of this project has been patented by KeyLemon.
More infos at the CTI official website.