Advancing Online Assessment Integrity: Integrated Misconduct Detection via Internet Protocol Analysis and Behavioral Classification

The rapid expansion of online learning has ushered in new educational opportunities, but concurrently introduced challenges in preserving academic integrity during assessments. This transition accentuates the need to address the heightened risks of e-cheating, encompassing various forms of academic dishonesty and potential misuse of advanced tools such as ChatGPT. To tackle these challenges, this paper proposes a pioneering multi-stage detection system that synergises geographical verification with behavioural analysis, offering a nuanced understanding of cheating behaviours in diverse online learning scenarios. The core innovation is an intelligent agent (multi-IA) incorporating an IP Detector for geographical verification and a deep learning-based Behavioural Monitor (known as DenseLSTM) for analysing response behaviours. This novel amalgamation enhances adaptability and accuracy in e-cheating detection, thereby fortifying the credibility and integrity of online assessments. Empirical analysis validates the multi-IA system’s efficacy, demonstrating remarkable accuracy in classifying candidate behaviours and effectively discerning between normal and potentially fraudulent responses. These findings underscore the system’s potential as an invaluable asset for educational institutions and educators in preserving the integrity of online assessments, and supporting the growth of online learning.