Home / Current Issue / Paper 1707762
Advanced Proctoring Framework for Exam Integrity
Subject area: Science,Engineering and Technology · Area of research: Deep Learning
Abstract
This research offers an intelligent invigilation system to maintain examination integrity by identifying unusual student behaviors through the use of deep learning. The model involves three phases: 1) verification of the student identity based on a face recognition method; 2) behavioral sampling to train the model, employing gesture analysis and convolutional 3D networks to analyze emotions; and 3) live video analysis of anomalous behavior, combining gesture and emotion analysis and student identification using face recognition. The model, trained on 4,000 training and 1,000 test images, classifies non-cheating activities with 99% accuracy and cheating activities with 97.6% accuracy. The suggested model performs better than other approaches, with accuracies of 98.4% for the detection of cheating behavior and 99.2% for non-cheating behavior, giving an overall accuracy of 98.8% and a low misclassification rate of 1.2%. Though the system exhibits strong accuracy, issues lie in scalability to larger classes with higher computational demands and requirements for more hardware for complete monitoring
Keywords
Suspicious Activity Detection, Exam Integrity, Deep Learning, Face and Gesture Recognition, Emotion Analysis
How to cite this paper
@article{1707762,
author = {Mithun S, Melvin M Shajan, Ragavendiran S, Sinduja K},
title = {Advanced Proctoring Framework for Exam Integrity},
journal = {Iconic Research And Engineering Journals},
year = {2025},
volume = {8},
number = {10},
pages = {118-129},
issn = {2456-8880},
url = {https://www.irejournals.com/formatedpaper/1707762.pdf},
abstract = {This research offers an intelligent invigilation system to maintain examination integrity by identifying unusual student behaviors through the use of deep learning. The model involves three phases: 1) verification of the student identity based on a face recognition method; 2) behavioral sampling to train the model, employing gesture analysis and convolutional 3D networks to analyze emotions; and 3) live video analysis of anomalous behavior, combining gesture and emotion analysis and student identification using face recognition. The model, trained on 4,000 training and 1,000 test images, classifies non-cheating activities with 99% accuracy and cheating activities with 97.6% accuracy. The suggested model performs better than other approaches, with accuracies of 98.4% for the detection of cheating behavior and 99.2% for non-cheating behavior, giving an overall accuracy of 98.8% and a low misclassification rate of 1.2%. Though the system exhibits strong accuracy, issues lie in scalability to larger classes with higher computational demands and requirements for more hardware for complete monitoring},
keywords = {Suspicious Activity Detection, Exam Integrity, Deep Learning, Face and Gesture Recognition, Emotion Analysis},
month = {April},
}