Cheating is the use of prohibited actions to illegally gain in the process of taking tests and exams. These actions cause negative consequences, making learners less capable of learning and creating, leading to unqualified individuals in the social workforce. In this study, we propose an action recognition method based on LSTM, VGG-19, Faster R-CNN, Mask R-CNN, and Vision Transformer architectures to classify fraudulent actions such as copying or pulling up to copy other students’ works, communicating, or transferring test material. Experimental results show that Vision Transformer identifies fraudulent actions with an accuracy of above 98%. This contributes to supporting teachers in managing candidates in exams, creating fairness and transparency in education.