Learning and examinations, thereby increasing the demand for automated cheating detection systems. This research aims to develop and process a representative dataset to detect cheating by analyzing eyebrow, eye, head, and hand movements using Convolutional Neural Network (CNN) and MediaPipe Hands. The study focuses on detecting movements such as "Furrowed Brow," "Look Up," "Look Down," "Normal," "Look Right," "Look Left," "Side Glance Left," "Side Glance Right," and "Raise Eyebrows." The results show that the model achieves an average accuracy of 76% in adequate lighting conditions (90 lux) but significantly decreases to 60.56% in low lighting conditions (20 lux). Lighting measurements were conducted in real-time using a lux meter. Despite the reduction in accuracy under low lighting, the model remained effective at predicting within a distance range of 30-70 cm between the subject and the camera. This research provides a foundation for further development of more accurate cheating detection systems across various lighting conditions.