Facial recognition models for proctoring report accuracies up to 99.21%, with limitations in dim-light and blurring conditions¶

Claim · All claims
Evidence · 1 study · review r2 · q2
Currency · newest study 2026 · not yet checked against its sources · not reviewed by a person · recheck by 2029: tool and model accuracy changes quickly

Subclaims¶

q2 i? The review reports facial recognition systems for proctoring achieving 99.21% (Elharrouss et al., MTCNN-based), 98.22% (YOLO Face) and 98.63% (MTCNN) accuracy, with noted limitations in dim-light conditions and blurring. → Malhotra 2026

Evidence¶

Malhotra 2026¶

Malhotra, M., & Chhabra, I. (2026). A comprehensive review of the changing landscape of academic dishonesty in automated proctoring in the era of artificial intelligence. Discover Education. https://doi.org/10.1007/s44217-026-01275-6

q2 · i? · review · r2

The review reports, citing Elharrouss et al., an MTCNN-based face recognition model evaluated on FLW, CEP and CASIA-WebFace datasets that "achieved an accuracy of 99.21%". Ganidisastra et al.'s CNN-FR approach is reported with YOLO Face at 98.22% and MTCNN at 98.63% accuracy.

"They utilized CNN for face recognition with the FLW, CEP , and CASIA-WebFace datasets and evaluated the model using the Pointing04 and Schneiderman datasets. Their model achieved an accuracy of 99.21%."

Discussion¶

Related Claims¶

• Eye-gaze-tracking-based proctoring systems report cheating-detection accuracies around 95–98% in their original studies — related
• Head-pose-estimation-based proctoring systems report widely varying accuracies, from 75.6% for cheating detection to 100% for yaw-angle estimation — related
• Deep learning techniques such as CNNs and RNNs better detect cheating from visual cues than traditional techniques, but each carries stated trade-offs — related
• Model accuracy of vision-based proctoring systems can vary across demographics, lighting conditions, and hardware setups, potentially disadvantaging some students — a broader claim this one bears on