Now a day’s facial expression recognition is becoming hot research for identifying the human mental state and his ability. As every individual try to express his feelings through his facial expression, it is becoming very difficult to identify the current state. In this proposed work, we use AI, expressly a deep neural network to figure out what are the potential results that a companion demand is trustworthy is or not. Each condition at every neuron (focus)is put through a Sigmoid cutoff. We utilize a plan enlightening record by Facebook or other social affiliations. This would permit the familiar huge learning calculation with get to know the examples of bot lead by back extension, confining the last expense work and changing every neuron's weight and propensity. Each data neuron would be a substitute, actually picked part of each profile changed over into a mathematical worth and if indispensable, bound by a discretionary number to limit one section merely affecting the outcome than the other. The proposed application try to split the human face into several parts and each and every part is assumed to be CNN and this will be undergo very precise examination and then come for conclusion. The CNN focuses on each middle point would be in danger for precisely one exceptional coordinated effort