This research paper defines a multi-modal system for verification based on the biometric fusion of retina, finger vein and finger print recognition. We have proposed feature extraction in retina recognition model by using SIFT and MINUTIA feature extract at work in different levels.Security is the main concept in ATM (Automated Teller Machines) today. Multi-modal Biometrics are secured as compared to uni-modal biometrics as even if single trait destroys the other is present. The application of multi-modal biometrics can be ATM. The proposed work, adds three biometric traits of a user namely retina, fingerprint and finger veins by an implemented software, later these are pre-processed and combined (Fused) together for score level fusion approach used. Retina is selected as a biometric trait as no binary retina feature matches unless they are of the similar user also retina has a good vessel pattern making it a good verifying approach as compared to other biometric traits. Security is found in the system by multi-modal biometric fusion of retina with finger vein and finger print. Feature Extraction approach and cryptography is used in-order to achieve security. The feature extraction is done with the help of MINUTIA and SIFT algorithm which are then classified using Deep Neural Network(DNN). The feature key points or minutiae points are fused at score level using distance average and later matched.The experimental result evaluated using MATLAB 2013a, illustrates the important enhancement in the performance of multi-modal biometric systems with higher values in GAR and FAR percentages.