Basic Study on Presentation Attacks against Biometric Authentication using Photoplethysmogram

Several kinds of biometric authentications have been used as countermeasures against identity spoofing. Recently, some approaches have utilized time-series biosignals for biometrics, and one of the approaches is photoplethysmogram (PPG)-based authentication. PPG sensing has the advantage of fewer restric-tion of measurement sites than other time-series physiological signals. Moreover, it can connect the authentication and the healthcare applications seamlessly with one sensor. However, identity spoofing against PPG-based authentication may occur by exploiting this advantage. To develop a PPG-based authentication system with countermeasures, we propose the feasibility of a presentation attack against PPG-based authentication. The attack stealthily records PPGs on non-genuine measurement sites, and transmits the signals to the authentication device, thereby utilizing the advantage of PPG sensing in which signals can be recorded on various sites of a subject ʼ s body. We conducted an experiment to investigate the feasibility of the attack. We developed a PPG-based authentication system comprising a PPG sensing system including PPG sensors for multiple measurement sites, and an authentication algorithm based on an existing PPG-based identification algorithm. We recorded PPGs on three measurement sites on the subjects ʼ bodies using the developed system. Then, we investigated the feasibility of the attack by inputting the feature values extracted from the PPGs recorded on non-genuine measurement sites to the classifier generated by the values from the PPGs recorded on genuine measurement sites. The results indicate that the attack can occur within a short time without any mapping under an ideal condition. Therefore, countermeasures such as liveness detection and utilization of unique information of measurement sites are required against the attack. implemented PPG-based authentication. The values are extracted from each PPG segment.