Design and Implementation of Authentication System Using Deep Convoluted Siamese Network

The majority essential prerequisite in this day is to conquer the various sorts of threats. Human behavioral and physiological components have the biggest alternative to overcome these security issues. In any case, the current biometric authentication methodology like fingerprints, faces and iris are profoundly complex methods. So in this paper, a new authentication system i.e. Finger vein authentication with the help of Deep Convoluted Siamese network has been proposed. The region of interest was done by threshold value over captured images by NIR and CCD cameras. After that, the Deep convoluted Siamese network is used to compare and contrast between two images to predict whether the two images are similar or not. The modern Siamese network uses a "Triplet Loss Function". In this triplet loss function, three fundamental images (Anchor Image, Positive Image and Negative Image) are considered where each image is a pixel matrix. The detection accuracy for testing and training is 97.2% and 96.4 % which is compared by utilizing several machine learning techniques (ANFIS, SVM, MLP and Global mapping/SVM) accuracy. It is clear from the comparative analysis that the proposed method gives better results than other algorithms. The proposed methodology TPR, FPR, FNR and TNR are 0.916, 0.027, 0.041 and 0.027 respectively. From these values, it is obvious that the proposed model gives better results, as the TP value is higher while the FP, FN and TN value are lower.