Examination is one of the traditional methods of assessment in India. During invigilation face recognition system was used to detect the students cheating activities. The traditional method takes more time, requires human resource and there is more chances of fault. The continuous monitoring is done by the exam supervisor for avoiding the malpractice in traditional method. The existing system can be replaced by deep learning method to improve the efficiency of automatic invigilation system. Deep learning is used to automate predictive analysis through which we can gain a lot in education sector. The few applications in education sector are language translation, chat bots and service bots. This paper discusses about web cam based invigilation system which is followed through the face detection and hand recognition methods. The main objective is to develop a model to detect mischievous activities among the students during real time examination in the classroom. This can be detected through object detection, hand movement recognition along with the face recognition. This model will lead to automatic invigilation system using deep learning approach.