Title: The Cyber Security Biometric Authentication based on Liveness Face-Iris Images and Deep Learning Classifier

Year of Publication: Aug - 2019
Page Numbers: 16-26
Authors: Sulaiman Alshebli, Mahmoud Shafik, Fatih Kurugollu
Conference Name: The 8th International Conference on Informatics and Applications (ICIA2019)
- Japan

Abstract:


This paper explains the liveness detection of the biometric system using the Face-Iris Images and deep learning classification. We have proposed novel hybrid algorithms for Face-Iris Liveness Recognition which can be used in cyber security authentication. In proposed model the individual identification is obtained from the extracted features that captured from face-Iris Images. Iris recognition is one of the most important biometric traits in which the iris image will be processed by some type of reliable, robust, and fast algorithm to capture the unique features embedded in iris. The camera system in which is used to capture the iris image will register depicts changes and variations in size of Iris as exposed to the light during the checkup and individual identification. This information will be used to verify the liveness of the iris image and distinguish the real lived iris image from the faked iris image. The recognition system based on the iris images are costly and required high resolution optical sensors and camera system beside the presence of the individual at the time of authentication and verification. The face image may change in certain degree such as appearance but shape and structure of face skeleton , location of eyes, mouth, and nose are remain unchanged. Due to this remarkable characteristic and ability to generate the face image features with near perfect identifiers and lack of requirement for physical contact with recognition system has made the facial recognition system remarkable. Our proposed face-iris feature extraction has been used to detect the fake face or iris from the real one and genuine live iris from the printed iris information by using the cosmetic lens. Our proposed model is based on the iris and face images in which high resolution camera and sensors are required at site to distinguish liveness Iris from fake one along with the high resolution algorithms that use combinations of DWT and SVD. The findings and results have shown the superiority our method, DWT-SVD, compared with hybrid of DCT, HT, MT with SVD, respectively. Our algorithm could be used to identify the fake Iris and face and greatly improve the cyber security. The extracted features can be obtained from videos or face images which have been captured from the scene even without individual physical contract. Our proposed model has been used for authentication and identification of the individuals based on the extracted features obtained from a new hybrid algorithm using both SVD and DWT. In proposed algorithm the face-iris features have been extracted and they are saved in the face-iris feature vectors for recognition. The Euclidian Distance vector has been used to classify the feature vectors extracted from live and fake face-iris images. The result and comparison of the proposed face-Iris recognition based on DWT-SVD with other technique has shown. Our method not only has higher speed of operation compared with other techniques but its rate of recognition is also good s as shown in the tabulated results. Experimental results also have shown that the hybrid methods of MT-SVD and HT-SVD have a better recognition results compared with DWT-SVD but their complex computation and extra time requirements for real time systems make them less important that DWT-SVD.