Authors: Manisha A. Nirgude, Sachin R. Gengaje Abstract:With an ever-increasing emphasis on security, the requirement of reliable methods for user authentication has amplified drastically and biometric methods for authentication are gaining wide acceptability. In contrast to the existing biometric features of human beings, iris, part of human eye has several interesting and unique characteristics. We solely address the iris characteristics and traits for human identification, explicitly for iris recognition. We have proposed Convolutional Neural Network (CNN) model to develop robust, scale & rotation invariant and scalable iris recognition system. Pretrained network model ‘Alexnet’ and data augmentation techniques viz. clockwise & anti-clockwise rotation, increased contrast, histogram equalization, addition of noise, dilation are used to create different datasets to train the CNN model. The performance of the proposed CNN model has been validated on five databases. We have achieved almost 100% iris recognition accuracy with reduced ERR as small as 0.46% as compared to existing iris recognition systems handling the issue of scale & rotation invariance and scalability using complete eye images.