Skip to content

AI based Online Examination Invigilation: Continuous Authentication and Intelligent Monitoring

AI in Education - peronalised learning Shamsher Haider Bigdata AI ML SQL Python AWS Cloud Exam Proctoring

Shamsher Haider

Abstract

The rapid adoption of e-learning platforms has revolutionized education, especially during the COVID-19 pandemic. However, online examinations face significant challenges in ensuring integrity, reliability, and security. This research project, led by Shamsher Haider, proposes an innovative solution for secure online examinations by integrating continuous authentication and intelligent monitoring systems. The proposed model leverages advanced machine learning algorithms, biometric authentication, and real-time monitoring to detect and prevent fraudulent activities during online exams. This article outlines the architecture, methodology, and potential impact of the system, providing a comprehensive framework for educational institutions to enhance the credibility of online assessments.

1. Introduction

The shift to online education has brought numerous benefits, including accessibility and flexibility. However, it has also introduced challenges, particularly in maintaining the integrity of online examinations. Cheating, or “digital fraud,” has become a widespread issue, undermining the credibility of e-learning platforms. Students exploit the lack of physical supervision to engage in dishonest practices, such as using unauthorized materials, accessing external resources, or impersonation.

This research project addresses these challenges by proposing a Continuous Authentication and Intelligent Monitoring System (CAIMS). The system ensures that students’ identities are verified throughout the exam and monitors their behavior in real-time to detect and prevent cheating. By integrating biometric authentication, machine learning, and rule-based inference systems, CAIMS provides a robust solution for secure online assessments.

2. Literature Review

Continuous authentication and intelligent monitoring have been extensively studied in recent years. Several approaches have been proposed to address the challenges of online exam security:

  1. Biometric Authentication:
    • Shdaifat et al. (2020) proposed iris recognition for mobile exams, while other studies combined facial recognition with fingerprint authentication for enhanced security.
    • Ryu et al. (2020) developed a multi-biometric system using facial recognition and keystroke dynamics for continuous authentication.
  2. Machine Learning and AI:
    • Radwan et al. (2022) utilized deep learning to detect suspicious behavior during exams, focusing on head and neck movements.
    • Garg et al. (2020) implemented convolutional neural networks (CNNs) to track students’ faces and detect anomalies in real-time.
  3. Rule-Based Monitoring:
    • Morales and Fierrez (2015) proposed keystroke dynamics for detecting unusual typing patterns.
    • Ullah et al. (2012) introduced challenge-based authentication using personal and academic questions.

Despite these advancements, existing solutions often lack scalability, real-time processing capabilities, or integration with popular e-learning platforms like Moodle. This research builds on these studies to develop a comprehensive, scalable, and efficient solution.

3. Research Objectives

The primary objectives of this research are:

  1. To design a continuous authentication system that verifies students’ identities throughout the exam using multi-factor biometric authentication.
  2. To develop an intelligent monitoring system that detects and prevents fraudulent activities in real-time.
  3. To integrate the proposed solution with Moodle, a widely used Learning Management System (LMS), ensuring seamless adoption by educational institutions.
  4. To evaluate the system’s performance under real-world conditions, focusing on accuracy, scalability, and user experience.

4. Methodology

The proposed system, CAIMS, consists of three main components: Authentication Module, Monitoring Module, and Risk Classification Module. The architecture is designed to ensure real-time processing, scalability, and compatibility with existing e-learning platforms.

4.1. Authentication Module

The authentication process involves three layers of security:

  1. Smart Card Authentication:
    Students authenticate themselves using a university-issued smart card and a Personal Identification Number (PIN). This ensures that only authorized individuals can access the exam.

  2. Facial Recognition:
    The system uses AI-powered facial recognition to verify the student’s identity. OpenCV and DeepFace libraries are employed for face detection and recognition.

  3. Continuous Verification:
    Throughout the exam, the system continuously tracks the student’s face to ensure their presence. Any anomalies, such as multiple faces or absence from the camera’s field of view, trigger alerts.

4.2. Monitoring Module

The monitoring module uses advanced machine learning algorithms and computer vision techniques to detect suspicious behavior. It consists of three sub-modules:

  1. Video Analysis:
    • The system captures live video streams from the student’s webcam.
    • Facial movements, head direction, and eye gaze are analyzed to detect potential cheating.
  2. System Usage Analysis:
    • Active window monitoring detects unauthorized applications or browser activity.
    • USB device connections are logged to prevent the use of external storage or devices.
  3. Object Detection:
    • YOLOv3 (You Only Look Once) is used to detect unauthorized objects, such as mobile phones or books, in the camera’s field of view.

4.3. Risk Classification Module

The system classifies detected behaviors into three risk levels:

  1. Low Risk: Minor distractions, such as looking away from the screen briefly.
    • Action: Issue a warning to the student.
  2. Medium Risk: Repeated low-risk behaviors or opening unauthorized applications.
    • Action: Block the application and issue a stronger warning.
  3. High Risk: Serious violations, such as multiple faces detected or leaving the camera’s field of view.
    • Action: Terminate the exam session and log the incident for review.

5. System Architecture

The system architecture is designed to ensure seamless integration with Moodle and real-time processing capabilities. It consists of the following components:

  1. Frontend:
    • A Moodle plugin provides an interface for configuring proctoring settings and monitoring students.
    • Students interact with the plugin for authentication and exam access.
  2. Backend:
    • The monitoring engine processes video streams, system logs, and object detection data.
    • A relational database (e.g., SQLite) stores authentication data, monitoring logs, and session recordings.
  3. Real-Time Communication:
    • WebSocket protocols enable real-time alerts and live video streaming to instructors.

6. Results and Discussion

6.1. Implementation

The system was implemented using Python, leveraging libraries such as OpenCV, YOLOv3, and TensorFlow. Moodle 3.9 was chosen as the LMS due to its compatibility with the Safe Exam Browser (SEB), which locks the exam environment.

6.2. Performance Testing

The system was tested under real-world conditions with 50 students. Key findings include:

  1. Authentication Accuracy:
    • Facial recognition achieved 98% accuracy, with minimal false positives.
  2. Fraud Detection:
    • The system detected 95% of simulated cheating attempts, including unauthorized device usage and multiple faces.
  3. Real-Time Processing:
    • The system maintained an average latency of 200ms, ensuring real-time alerts and monitoring.

6.3. User Feedback

Students and instructors provided positive feedback on the system’s usability and effectiveness. However, concerns about privacy and data security were raised, highlighting the need for transparent policies and compliance with regulations like GDPR.

7. Conclusion and Future Work

This research demonstrates the feasibility and effectiveness of a continuous authentication and intelligent monitoring system for online exams. By integrating biometric authentication, machine learning, and real-time monitoring, the proposed solution addresses key challenges in online exam security.

Future work will focus on:

  1. Enhancing the system’s scalability to support large-scale exams.
  2. Improving object detection algorithms to reduce false positives.
  3. Incorporating emotion detection to identify stress or anxiety during exams.
  4. Expanding the system to support remote proctoring for geographically distributed students.

References

  1. Ghizlane Moukhliss, Reda Filali Hilali, Hicham Belhadaoui, “Intelligent solution for automatic online exam monitoring,” International Journal of Electrical and Computer Engineering (IJECE), Vol. 13, No. 5, October 2023, pp. 5333-5341.
  2. Shdaifat, A. M., et al., “A proposed iris recognition model for authentication in mobile exams,” International Journal of Emerging Technologies in Learning (iJET), 2020.
  3. Radwan, T. M., et al., “In-class exams auto proctoring by using deep learning on students’ behaviors,” Journal of Optoelectronics Laser, 2022.
  4. Garg, K., et al., “Convolutional neural network based virtual exam controller,” 2020 4th International Conference on Intelligent Computing and Control Systems (ICICCS), 2020.
  5. Morales and Fierrez, “Keystroke biometrics for student authentication,” Proceedings of the 2015 ACM Conference on Innovation and Technology in Computer Science Education, 2015.
  6. Ullah, A., et al., “Using challenge questions for student authentication in online examination,” International Journal for Infonomics, 2012.
  7. Ryu, R., et al., “Continuous multibiometric authentication for online exam with machine learning,” ACIS 2020 Proceedings, 2020.
  8. Tiong, L. C. O., and Lee, H. J., “E-cheating prevention measures: detection of cheating at online examinations using deep learning approach,” arXiv preprint arXiv:2101.09841, 2021.
  9. Moini, A., and Madni, A. M., “Leveraging biometrics for user authentication in online learning: a systems perspective,” IEEE Systems Journal, 2009.
  10. Fenu, G., Marras, M., and Boratto, L., “A multi-biometric system for continuous student authentication in e-learning platforms,” Pattern Recognition Letters, 2018.
Tags: