Face recognition is one of the most common biometric identification techniques. Despite several techniques being proposed, face recognition research on powerful deep models like Siamese networks remains less researched. Siamese networks are not as deep as traditional CNNs and offer unique advantages like small memory footprint, lesser training time, and overall network simplicity which could open up new pathways for face recognition. In this paper, the distance function of the Siamese network has been modified and promising results have been obtained. Face recognition on the popular LFW database using the proposed technique increases the state-of-the-art accuracy by 0.28%.