The procedure of processing a picture, obtaining information, and evaluating the information to extract domain-specific knowledge is known as data analytics. The present development involves not only evaluating any photograph to determine data retrieval, but also evaluating live surveillance films to identify events occurring within its coverage area. These systems are going to be put into use immediately. Employing a training model like an artificial neural network makes automated face recognition from security footage easier. Estimating skin color aids in hand detection. This research project uses surveillance camera footage from exams to identify suspicious activity, including object exchange, a fresh entry, looking through someone else’s answer sheet, and person transfer. These days, people are more concerned with the impartiality of monitoring, thus it is important to identify unusual conduct so as to maintain the integrity of criminal identification. The majority of approaches used now suggest models for specific abnormal conduct. We identify the optical flow of the video data in this framework and suggest a 3D convolution neural network model to address the issue. This calls for the use of facial detection, criminal detection, and the ability to distinguish among the faces of various people. The rate of mistake caused by human oversight will be reduced with the automation of “criminal identifying & detection”.