Search In this Thesis
   Search In this Thesis  
العنوان
Automatic Liveness Detection from Facial Images /
المؤلف
El-Sayed, Mehad Araby Hassan.
هيئة الاعداد
باحث / Mehad Araby Hassan El-Sayed
مشرف / Ayman Mohamed Wahba
مشرف / Mohamed Nabil Moustafa
مناقش / Mohamed Nabil Moustafa
تاريخ النشر
2019.
عدد الصفحات
117p. :
اللغة
الإنجليزية
الدرجة
ماجستير
التخصص
الهندسة الكهربائية والالكترونية
تاريخ الإجازة
1/1/2019
مكان الإجازة
جامعة عين شمس - كلية الهندسة - هندسة كهربائية
الفهرس
Only 14 pages are availabe for public view

from 117

from 117

Abstract

Liveness detection is a vital research field when using biometric systems in securing our data. As with technology increasing the usage of the face is preferred in protecting mobile and laptops data. But most devices use biometric systems that are easily attacked by using images of the user, 3D masks, HD videos or some makeup. Thus, the need for liveness detection is increasingly required. As it can detect if the live person in front of the camera in the capture time or it’s spoofed one.
Liveness Detection can be implemented using several ways; as using special cameras, detecting some user liveness by asking the user some questions, asking the user to make some movements by head or hand, or by detecting face features as we perform in our thesis.
We enhanced the OULU approach that detects the texture feature and local shape analysis of the face by replacing HOG with customized SIFT to detect the local feature and as for texture, we used LBP and Gabor wavelet.
We perform our tests first in replay attack dataset then get the best parameters that are applied to another 2 publicity datasets NUAA and CASIA. The results in the three datasets are better than
VIII
OULU approach. As We use EER and ROC curves to show the experiment results and comparison between datasets.
By applying the proposed approach AUC increases from 0.9741 to 0.9764 NUAA, and on CASIA the area under the curve (AUC) increases from 0.9925 to 0.9974 and on Replay Attack the AUC increased from 0.9452 to 0.9603 And EER from 7% to 6.25% .
Keywords: Images, Liveness, Faces, Detection,
Data-set, and spoofing