Digital Signal Processing and Machine Learning for Exam Fraud Detection

Abstract

One of the main issues that presents difficulty in online learning settings is academic dishonesty. In these environments, traditional proctoring techniques are not always successful; hence, a new hybrid system is under development to address the problem. This creative system detects cheaters in the act by combining digital signal processing and image processing. Multiple cameras in the image processing component of the system monitor students during tests and flag any suspicious activity, including too forceful hand gestures or head movements. Simultaneously, the digital signal processing element finds any illegal devices and blocks internet access to stop digital cheating. Under tests, the system produced some rather outstanding results. It was 100% successful in blocking internet access and able to correctly identify illegal devices 96% of the time and aberrant physical movements 80% of the time. Over current approaches that only use one of these techniques, the twin approach of tracking behaviour and identifying devices represents a significant improvement. The system still has certain constraints and is not flawless, though. The accuracy rate falls to 65% in noisy or complicated surroundings, which emphasises the need for more improvement. Schools and colleges wishing to put sensible plans into action to stop cheating and uphold academic integrity will find great use for this study.