The examination evaluates the learners’ ability to achieve a specific goal, representing a crucial assessment of the knowledge acquired during the learning process. To attain success in the examination, it is imperative to prevent cheating in the examination hall. Despite this, cheating still occurs due to the limited availability of human resources. Therefore, we have collected surveillance videos of examination halls from a high school in Vietnam to analyze and implement deep learning architectures such as You Only Look Once (YOLO) and Single Shot Detector (SSD) MobileNet V2 to detect anomalous behavior among students during the examination. Our study focuses on detecting five common abnormal behaviors, including looking around, bending over the desk, putting one or two hands under the table, waving, and standing up. YOLO achieved the best results, with a performance of 83.55%, 99.65%, 97%, 99.2%, and 98.0866% in Intersection over Union (IoU), Mean Average Precision (mAP), Precision, Recall, and F1-Score, respectively, across 2639 images. This approach is expected to assist educators and teachers in detecting and preventing cheating activities in examination rooms.