An Ear Anti-Spoofing Database with Various Attacks
Jalil Nourmohammadi Khiarak, Andrzej Pacut · 2018
The Biometrics of the ears have both advantages and disadvantages compared to other physical attributes. The small surface and the relatively simple structure have a controversial effect. In a positive way, these features provide faster processing compared to face detection and make detection easier compared to fingerprints. On the other side, like other biometrics, current ear biometric recognition systems are vulnerable to attacks. A spoofing attack occurs at sensor level and every impostor can masquerade as someone else by altering data, thus, obtaining an illegitimate access. Due to a lack of anti-spoofing databases, that would support this paper, ear fake databases have been built using different mobile phones. In this paper, an ear presentation attack detection database is collected which contains a various range of variations of potential attacks. In particular, the database consists of two main parts, a) AMI dataset which has 700 ear images and we make display attack by using them, b) data collected at University of Tabriz containing 20 genuine subjects and fake ears which are made from the genuine ears. Different mobile phones are used for collecting the database. Three fake ear attacks are implemented which include video attack, printed attack, and display attack. Consequently, for each subject, 2 videos (left and right ears), 8 different images, and the final database contain 10 video clips and 160 images are prepared. General Image Quality Assessment is used as a baseline algorithm for comparison which is used vastly in the liveness detection purpose. Releasing the first database in ear liveness detection can open new ways for investigating on ear biometrics systems more confidently to use future research on mobile smartphones.