Anti-cheat methods for face authentication system spoofed by high resolution fake face

• Main Authors: Chiu-Yuan Tai, 戴久芫
• Other Authors: Chin-Lun Lai
• Format: Others
• Language: zh-TW
• Published: 2015
• Online Access: http://ndltd.ncl.edu.tw/handle/jxte5t

碩士 === 亞東技術學院 === 資訊與通訊工程研究所 === 103 === Due to the multiple convenient qualities (quick, remote detection ability, non-contact), face detection has been widely applied in fields such as access control, monitoring, auto focusing systems, or verification of the subject’s identity and behavior. However, with the widespread adoption of face detection technique, spoofing technique had become increasingly advanced with face Information being forged or collected to deceive or bypass the verification of face detection system. Thus, it is crucial for the biometric system to identify the forged characteristics. In this thesis, three fake face detection strategies based on analyzing characteristics of high-definition display, including the brightness of face image, DCT component of face edge, and hue of face image, are proposed to detect the capturing images. By using a PNN model to establish the image analysis system, the system can effectively identify fake face images from the input sequences. Moreover, to reduce the error probability from a single shut, sequence analysis method is also proposed to improve the system identification correctness, thus increase the stability and practicability of the proposed system.