Identity Verification in Real Time Proctoring: An Integrated Approach with Face Recognition and Eye Tracking

Online courses and home-based work create an urgent need for effective real-time online proctoring services with reliable identity authentication. Traditional solutions struggle with issues such as bias and under-representation of certain demographics in facial recognition, which compromises the validity of online exams. This research describes a new solution that combines Siamese network architecture with a state-of-the-art Inception ResNet v1 model. Transfer learning from the VGGFace2 dataset is used to enhance the network’s ability to identify small but critical facial characteristic markers. Additionally, the solution integrates an eye-tracking methodology that uses machine learning algorithms for attention analysis. This method significantly improves the detection accuracy of minor face differences among different demographics, which is critical for reliable face recognition. The eye-tracking feature focuses on modeling applicants’ behaviors (attention, intent, etc.) for a deeper understanding of examinees’ actions in online exams for monitoring and academic integrity purposes. The above models collectively produce a powerful solution for institutions with capabilities far exceeding the limitations of existing compute and financial resources. This research focuses on the feasibility of an online proctoring solution that is unbiased, effective, and dynamic to adapt to evolving demands of online education and professional environments around the world.