In our society, sometime we hide our genuine feeling and emotion and purposely express different emotion in front of our surrounding folks. But as it’s not actually a natural emotion, hence, it is more or less, predictable by others. Human vision system has enormous capability to recognizing genuine and fake smile of an individual. Discriminating genuine and fake smile is very thought-provoking task and even though very smaller amount of research has been carried out in this topic. In this paper, we are exploring a method to distinguish real from fake smile with high precision by using convolution neural networks (CNN). System has been train with FERC-2013 dataset having seven types of emotions namely happy, sad, disgust, angry, fearful, surprised and neutral. Emotions percentages of real and fake face are recorded by the emotion detection system. Based on recorded score, we investigate the effect of various percentages of emotions presented on both faces and then we are going to classify the smile on the face is real or fake.