Abstract
Online proctoring is a monitoring tool that instantly recognizes any unlawful behavior both before and during the exam. Three different sources are used for surveillance: audio, video streams, and desktop screen record-ings. In general, a proctor, a specifically trained professional who checks to see if the system has made a mistake and looks for violations from students taking the exam, always keeps an eye on the proctoring process. Yet, it is impossi-ble for one individual to monitor more than two people at once, and it is also possible to let fraudulent students sit for the test. Determining how an online proctoring system will be very successful in preventing unfair testing by employ-ing face recognition on video samples is the focus of this paper. The development of a trainable artificial intelligence network makes it possible to employ multiple variables simultaneously. We developed a deep learning technique for automatic face identification as part of this work, employing a convolutional neural network to calculate the differ-ence between a sample and a face in a video. It was typical to employ video samples of various qualities, taken at various distances, from various angles, with various lighting conditions, and with the inclusion of accessories to get more accurate findings. The simulation results showed that, with the exception of the experiment when accessories such as a medical face mask were present, the model we trained accurately predicted the outcome for each sampling criterion. Further examples show the effectiveness of face recognition, also in the quick playback format, the percentage of face recognition reached the minimum threshold (slightly more than 75 %).