In many developing countries, only about 30% of eligible candidates win a university seat, and job recruitment exams are just as competitive. Some candidates send a “proxy” to sit the exam for them. This research proposed a fast face recognition system to stop that.
The problem
With huge numbers of candidates, exam administrators cannot check every person individually. Altered photos on registration papers make verification even harder. That weakness lets some candidates send better-prepared proxies in their place.
What the research did
We proposed an automatic examinee validation system using eigenfaces, a classic face recognition technique that represents faces using a small set of characteristic patterns. It compares the person at the exam centre with the registered candidate.
What we found
The system was fast, efficient, accurate and simple to implement, designed to make it almost impossible for a proxy examinee to enter the exam centre.
Who benefits
- Admission and recruitment test organisers
- Honest candidates, who deserve a fair contest
- Education systems in developing countries
Limits and what comes next
Eigenfaces is an early technique; modern deep learning methods are more robust to lighting and pose. Any face recognition system needs safeguards for privacy and to avoid wrongly rejecting genuine candidates.
The team
I led this research with Mohammad Abu Yousuf and Md. Fazlul Karim Patwary (Jahangirnagar University).
Frequently asked questions
What is a proxy examinee?
Someone who sits an exam in place of the registered candidate.
What are eigenfaces?
A face recognition method that represents faces as combinations of a few characteristic patterns.
Is face recognition fair?
It must be tested carefully across different people and used with privacy safeguards and a human fallback.