April 9, 2021
MSU research on face spoof detection earns international honor
Biometrics research on face spoof detection at Michigan State University has received international acclaim.
The Computer Science and Engineering (CSE) team of University Distinguished Professor Anil Jain and former postdoctoral associates Hu Han and Di Wen, has been awarded the Best Paper Award from the IEEE Signal Processing Society (SPS). Only three Best Paper honors were selected for recognition among the society’s nine different peer-reviewed magazines and transactions in 2020.
IEEE SPS has been the world’s premier professional society for signal processing scientists and professionals since 1948. It has more than 19,000 members that span across 100 countries worldwide.
The paper receiving the award, "Face Spoof Detection With Image Distortion Analysis," was published in the IEEE Transactions on Information Forensics and Security. According to Google Scholar, the paper has already received 449 citations.
With the wide deployment of face recognition systems in applications ranging from border control to mobile device unlocking, solutions to combat face spoofing attacks requires increased attention because such attacks can be easily launched via printed photos, video replays, and 3D masks. Failure to detect face spoofs can be a major security threat and can seriously compromise user privacy.
Jain said MSU’s face spoof detection algorithm is robust given its use of four different features -- specular reflection, blurriness, chromatic moment, and color diversity. An ensemble classifier, trained for different types of face spoof attacks, is used to distinguish between genuine (live) and spoof faces. The approach extends to face spoof detection in videos using a voting-based scheme on top of multiple video frames. As part of the study, the researchers collected a face spoof database, called MSU mobile face spoofing database (MSU MFSD), using two mobile devices (Google Nexus 5 and MacBook Air) with three types of spoof attacks (printed photo, replayed video with iPhone 5S, and replayed video with iPad Air).
“Our results highlight the difficulty in separating genuine and spoof faces, especially in cross-database (different collection protocols for training and test face datasets) and cross-device (different cameras for collecting training and test face images) scenarios,” Jain added.
Han is now an associate professor at the Institute of Computing Technology, Chinese Academy of Sciences, Beijing. Wen is now a staff software engineer at LinkedIn.
