A proposed framework for vision based gait biometric system against spoofing attacks

Gait recognition based person identification through visual analysis is preferred over other biometric because gait is the only biometric trait that is perceivable from remote distance, extractable from low quality visuals and did not require subject's cooperation for data collection. Extensive research in the field of vision based gait recognition have enable researchers to achieve milestones of person identification, age estimation and gender recognition through visual analysis of gait. Moreover, rapid growth of security camera networks and sparse implementation of visual analysis on security visuals urges researchers to design vision based gait recognition systems for person identification. Person identification based on gait recognition can be implemented for monitoring of abnormal activities at public places and for access control of security critical places. Gait biometric based access control systems are equally vulnerable to spoofing attacks despite of passive and unnoticeable mode of action. In order to address limitations of vision based gait biometrics system under spoofing attacks, this paper proposed a methodology for detection of targeted spoofing attacks with different scenarios. The main contribution is to figure out replication of clothing, physical structure and gait dynamics can affect classification result. Since spoofing attacks against gait biometric systems are less explored, testing strategies are also suggested in the proposed work.