In many real world applications, multimodal biometric systems are popularly used due to its ability to deal with disadvantages of unimodal biometric systems. In this paper, effective and efficient multimodal biometric system are developed using Convolution Neural Network (CNN) which helps to predict identity of a individual based on his/her iris pattern. The proposed iris recognition systems is a combination of CNN with Soft Max classifier. The performance of the proposed iris is evaluated on IITD database obtained in various conditions. The obtained results shows that the proposed system outperforms the previous approaches. And the four layered CNN with Adadelta optimizer has produced the best results of more than 95% accuracy.