Abstract:
In many applications in recent days, facial detection is being widely used as a measure of
security feature in these applications. However, even with many robust facial detection
techniques, these systems can be easily spoofed with a use of printed picture or face video. The
proposed work involves the development of a robust system using ultrasonic sensors to detect a
printed picture or face video spoof attack. Three ultrasonic sensors are strategically placed in
order to measure the data received by these sensors and compare each data in different scenarios.
In each scenario of live face, video and paper face, series of continuous data received from the
sensors is collected on which first degree and second degree quadratic regression is performed.
Results of these regression analyses determine a spoof attach against each set of collected data.
Final decision of real or spoofed face is determined by the majority count method. The results
achieved from the analysis of the data, clearly distinguish real face from paper or video face. The
proposed work opens up the possibility of many applications to correctly detect spoof attack and
hence increasing the security of applications.
