Reproduce “Cheating Detection Pipeline for Online Interviews and Exams” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Azmi Can Özgen, Mahiye Uluyağmur Öztürk, Umut Bayraktar

A fully automated pipeline identifies common cheating activities in online interviews and exams, such as the presence of another person, electronic device usage, and candidate absence, by analyzing standard webcam video streams. This system processes video faster than real-time (4.9 FPS for 3 FPS input), achieving an overall F1 score of 0.88 on a proprietary dataset.

Key insights

  • Processing videos at 3 frames per second is sufficient to extract meaningful cheating relations while significantly improving processing speed.
  • A novel 'partial face matching' strategy using the Frobenius norm effectively reduces false alarms by ignoring less significant features in face encoding comparisons.
  • The pipeline's modular design allows for flexible deployment, enabling components to be customized or detached based on specific application needs, enhancing its practical utility.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.