Online proctoring systems commonly make visual judgments frame by frame. In practice, however, changes in participant appearance or acquisition conditions can make ordinary posture shifts look suspicious, leading to brief false alarms or repeated warnings. This paper studies real-time suspicious-event monitoring with a role-separated design: the visual stream is kept as interpretable evidence for monitoring and review, whereas automatic risk estimation is handled by a lightweight causal acoustic path. YuNet, 13 pose-geometric features, 20 contour features, and a frozen Random Forest produce the visual score. On the audio side, 13 lightweight features are computed from a strictly causal trailing window, standardized, and evaluated by logistic regression; a persistence rule then converts the binary outputs into Suspicious Events. Experiments on the public OEP dataset use participant-level outer validation, with development, outer-test, and system validation evidence kept separate. In the clean outer test, Acoustic increases Macro Participant PR-AUC from 0.2987 to 0.5666 and Event F1 from 0.2003 to 0.4120, while FP/min decreases from 2.1604 to 1.0004 relative to Visual. The simple Visual + Acoustic setting reaches an Event F1 of 0.4050 and does not provide a stable additional gain. These findings motivate the final division of roles: acoustic cues drive automatic risk discrimination, while visual cues remain available for monitoring and human review. Across 250,932 streaming ticks, no binary, temporal-state, or alarm mismatches are observed; the core acoustic alarm engine has a P95 latency below 1 ms and does not require a GPU.

Tianzhen Chen. Real-Time Suspicious-Event Monitoring for Exam Integrity Using Visual Behavior Monitoring and Lightweight Acoustic Assistance. Academic Journal of Engineering and Technology Science (2026), Vol. 9, Issue 4: 55-62. https://doi.org/10.25236/AJETS.2026.090407.

[1] Y. Atoum, L. Chen, A. X. Liu, S. D. H. Hsu, and X. Liu. Automated Online Exam Proctoring[J]. IEEE Transactions on Multimedia, 2017, 19(7): 1609-1624. DOI: 10.1109/TMM.2017.2656064.

[2] K. Chen, E. Gal, H. Yan, and H. Li. Domain Generalization with Small Data[J]. International Journal of Computer Vision, 2024, 132(8): 3172-3190. DOI: 10.1007/s11263-024-02028-4.

[3] Q. Yang, C. Wang, P. Liu, Z. Jiang, and J. Li. Video Anomaly Detection via self-supervised and spatio-temporal proxy tasks learning[J]. Pattern Recognition, 2025, 158: 111021. DOI: 10.1016/j.patcog.2024.111021.

[4] P. Tejaswi, S. Venkatramaphanikumar, and K. Venkata Krishna Kishore. Proctor Net: An AI framework for suspicious activity detection in online proctored examinations[J]. Measurement, 2023, 206: 112266. DOI: 10.1016/j.measurement.2022.112266.

[5] T. Singh, R. R. Nair, T. Babu, and P. Duraisamy. Enhancing Academic Integrity in Online Assessments: Introducing an Effective Online Exam Proctoring Model using YOLO[C]//Proceedings of the International Conference on Machine Learning and Data Engineering (ICMLDE 2023). Procedia Computer Science, 2024, 235: 1399-1408. DOI: 10.1016/j.procs.2024.04.131.

[6] S. Essahraui, I. Lamaakal, Y. Maleh, K. El Makkaoui, M. F. Bouami, I. Ouahbi, et al. Deep Learning Models for Detecting Cheating in Online Exams[J]. Computers, Materials & Continua, 2025, 85(2): 3151-3183. DOI: 10.32604/cmc.2025.067359.

[7] Y. Liu, J. Ren, J. Xu, X. Bai, R. Kaur, and F. Xia. Multiple Instance Learning for Cheating Detection and Localization in Online Examinations[J]. IEEE Transactions on Cognitive and Developmental Systems, 2024, 16(4): 1315-1326. DOI: 10.1109/TCDS.2024.3349705.

[8] M. Fauss, X. Liu, C. Li, I. Choi, and H. V. Poor. Bayesian Selection Policies for Human-in-the-Loop Anomaly Detectors with Applications in Test Security[J]. Psychometrika, 2026, 91(1): 279-311. DOI: 10.1017/psy.2025.10056.

[9] R. Delussu, L. Putzu, and G. Fumera. Synthetic Data for Video Surveillance Applications of Computer Vision: A Review[J]. International Journal of Computer Vision, 2024, 132: 4473-4509. DOI: 10.1007/s11263-024-02102-x.

[10] S. Kaddoura and A. Gumaei. Towards effective and efficient online exam systems using deep learning-based cheating detection approach[J]. Intelligent Systems with Applications, 2022, 16: 200153. DOI: 10.1016/j.iswa.2022.200153.

[11] C. Li, H. Li, and G. Zhang. Cross-modality integration framework with prediction, perception and discrimination for video anomaly detection[J]. Neural Networks, 2024, 172: 106138. DOI: 10.1016/j.neunet.2024.106138.

[12] Y. Su, Y. Tan, S. An, M. Xing, and Z. Feng. Semantic-driven dual consistency learning for weakly supervised video anomaly detection[J]. Pattern Recognition, 2025, 157: 110898. DOI: 10.1016/j.patcog.2024.110898.

[13] D. P. M. Huy, G. N. Nguyen, and D.-N. Le. A Hybrid Deep Learning Approach for Real-Time Cheating Behaviour Detection in Online Exams Using Video Captured Analysis[J]. Computers, Materials & Continua, 2026, 86(3): 48. DOI: 10.32604/cmc.2025.070948.

[14] C. Park, D. Kim, M. A. Cho, M. Kim, M. Lee, S. Park, and S. Lee. Fast video anomaly detection via context-aware shortcut exploration and abnormal feature distance learning[J]. Pattern Recognition, 2025, 157: 110877. DOI: 10.1016/j.patcog.2024.110877.

[15] Y. Kim and Y.-G. Kim. MPE: Multi-frame prediction error-based video anomaly detection framework for robust anomaly inference[J]. Pattern Recognition, 2025, 164: 111595. DOI: 10.1016/j.patcog.2025.111595.