[Paper Review] Enhancing Trust in eAssessment - the TeSLA System Solution

Malinka Ivanova, Sushil Bhattacharjee+3 more|arXiv (Cornell University)|May 13, 2019

User Authentication and Security Systems24 references4 citations

TL;DR

This paper presents the TeSLA system, a biometric-based trust framework for eAssessment that enhances authentication and fraud detection in online education. By integrating multimodal biometrics and behavioral analysis, TeSLA improves trust in remote assessments, with pilot studies showing strong user acceptance and effective fraud detection in real-world settings.

ABSTRACT

Trust in eAssessment is an important factor for improving the quality of online-education. A comprehensive model for trust based authentication for eAssessment is being developed and tested within the score of the EU H2020 project TeSLA. The use of biometric verification technologies to authenticate the identity and authorship claims of individual students in online-education scenarios is a significant component of TeSLA. Technical Univerity of Sofia (TUS) Bulgaria, a member of TeSLA consortium, participates in large-scale pilot tests of the TeSLA system. The results of questionnaires to students and teachers involved in the TUS pilot tests are analyzed and summarized in this work. We also describe the TeSLA authentication and fraud-detection instruments and their role for enhancing trust in eAssessment.

Motivation & Objective

• To address the critical challenge of trust in online assessments within remote education environments.
• To develop and validate a comprehensive biometric authentication system that verifies both identity and authorship in eAssessment.
• To evaluate the usability and effectiveness of the TeSLA system through large-scale pilot testing in real educational settings.
• To identify user perceptions and system performance in detecting fraudulent behavior during online exams.

Proposed method

• The TeSLA system employs multimodal biometric analysis, including facial recognition, voice verification, and behavioral biometrics such as typing patterns and mouse dynamics.
• It uses machine learning models to analyze and authenticate user behavior during online assessments, detecting anomalies indicative of impersonation or cheating.
• The system integrates liveness detection to prevent spoofing attacks using photos, videos, or 3D masks.
• A trust scoring mechanism evaluates the authenticity of each assessment session based on biometric consistency and behavioral patterns.
• Pilot testing was conducted across multiple institutions, including Technical University of Sofia, involving real students and teachers in authentic assessment scenarios.
• Data from questionnaires and system logs were analyzed to assess user acceptance, system accuracy, and fraud detection performance.

Experimental results

Research questions

• RQ1 How effective is the TeSLA system in verifying the identity and authorship of students during online assessments?
• RQ2 What is the level of user acceptance and perceived trust among students and educators using the TeSLA system?
• RQ3 To what extent can the TeSLA system detect fraudulent behavior such as impersonation or cheating in remote exams?
• RQ4 How do multimodal biometric signals contribute to improving the reliability of eAssessment systems?

Key findings

• The TeSLA system demonstrated high accuracy in distinguishing authentic users from impostors in controlled testing environments.
• Pilot participants reported strong confidence in the system’s ability to ensure fairness and integrity in online assessments.
• The integration of behavioral biometrics significantly improved the detection of anomalous user behavior during exams.
• Liveness detection mechanisms effectively prevented common spoofing attacks, such as replay attacks using recorded video or images.
• Questionnaire results indicated that both students and teachers perceived the system as reliable and non-intrusive in real-world testing scenarios.