EUVIMEDEUVIMED entdecken
Crossref

Enhancing Examination Integrity through AI-Based Privacy-Preserving Proctoring for Computer-Based Testing in Nigeria

Sadiya Mohammed Dala, Mohammed Mansur Ibrahim

International Journal of Innovation in Science and Technology (IJIST) · 30.05.2026 · 1 (1) · 258-271

DOI: 10.64290/ijsat.v1i1.18 ↗

Kurz gefasst

Fachartikel in der Zeitschrift International Journal of Innovation in Science and Technology (IJIST), online erschienen am 30.05.2026. Verfasst von Sadiya Mohammed Dala und 1 weiteren Autor:innen.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Examination malpractice in Computer-Based Testing (CBT) environments remains a critical challenge in Nigerian federal universities, where bandwidth constraints and equity concerns preclude the use of cloud-dependent commercial proctoring solutions. This study presents, implements, and evaluates a privacy-preserving, multi-modal AI proctoring architecture that fuses acoustic anomaly detection, rule-based head-pose classification, and multi-face detection using a decision-level OR fusion strategy. A Bidirectional Long Short-Term Memory (BiLSTM) classifier was trained on publicly available Google Speech Commands and ESC-50 datasets; a deterministic head-pose detector was applied to 240 synthesised pose traces; and an OpenCV-based multi-face module flagged simultaneous candidate presence. All three suspicion signals were fused via a conservative OR rule under simulation conditions representative of the Federal University of Kashere (FUK) CBT infrastructure. The BiLSTM audio classifier achieved 97.92% accuracy, 97.5% precision, 98.3% recall, 97.9% F1, and 99.8% AUC-ROC. Decision-level OR fusion yielded 98.3% overall accuracy, 98.9% precision, 98.9% recall, 98.9% F1, and 99.0% AUC-ROC, with end-to-end latency of approximately 135 ms per window, well within the 250 ms non-functional budget. The proposed architecture demonstrates that a lightweight, on-device proctoring system can achieve near-commercial detection performance without transmitting biometric data to external servers, addressing both infrastructure and privacy constraints in a Nigerian university context. Recommendations include initiating a shadow-mode institutional pilot at FUK, developing a national AI proctoring standard through the National Universities Commission (NUC), and embedding AI proctoring governance within institutional ICT policy frameworks.

Zitieren

APA

Dala, S. M., & Ibrahim, M. M. (2026). Enhancing Examination Integrity through AI-Based Privacy-Preserving Proctoring for Computer-Based Testing in Nigeria. International Journal of Innovation in Science and Technology (IJIST), 1(1), 258-271. https://doi.org/10.64290/ijsat.v1i1.18

Vancouver

Dala SM, Ibrahim MM. Enhancing Examination Integrity through AI-Based Privacy-Preserving Proctoring for Computer-Based Testing in Nigeria. International Journal of Innovation in Science and Technology (IJIST). 2026;1(1):258-271. doi:10.64290/ijsat.v1i1.18

BibTeX
@article{Dala2026,
  author = {Dala, Sadiya Mohammed and Ibrahim, Mohammed Mansur},
  title = {Enhancing Examination Integrity through AI-Based Privacy-Preserving Proctoring for Computer-Based Testing in Nigeria},
  journal = {International Journal of Innovation in Science and Technology (IJIST)},
  year = {2026},
  volume = {1},
  number = {1},
  pages = {258--271},
  doi = {10.64290/ijsat.v1i1.18}
}
Kostenlos starten →