A Comprehensive Review of the Changing Landscape of Academic Dishonesty in Automated Proctoring in the Era of Artificial Intelligence

Synthesis: Malhotra & Chhabra (2026) comprehensively review the landscape of academic dishonesty across conventional, online, automated, and AI-based proctoring systems, catalyzed by COVID-19's shift to remote exams. They trace the transition from offline invigilation to online and AI/ML proctoring, document the cheating methods AI systems must counter (identity spoofing, browser/device use, copy-paste), and identify the practical, infrastructure, and psychological challenges — anxiety, lack of proficiency, cost, and connectivity — that shape proctoring adoption and effectiveness.

Key Findings

• Widespread cheating motivates proctoring. Studies cited report that ~37.8% of college and ~41.8% of high-school students admit to cheating or misconduct during exams — a key driver of monitoring investment.
• Transition from offline to online/AI proctoring. COVID-19 accelerated the shift from physical invigilation to online proctoring (human proctors via webcam/control center) and then to AI/ML-driven automated systems; common platforms include ProctorU and Kryterion.
• Cheating methods AI must counter. Identity spoofing (masking a face via photographs or video to impersonate another), browser/tab use to search online, and copy-paste from books, phones, or cheat sheets.
• Challenges of online proctoring. Test-taker anxiety (especially for users not proficient with online tools), proctor/test-taker proficiency gaps causing false malpractice accusations, and infrastructure requirements (webcams, microphones, internet) that are not affordable or available to all.
• AI/ML proctoring captures facial expressions and emotions through imaging systems, enabling richer behavioral analysis but raising cost and infrastructure demands.

Study Design & Method

A comprehensive thematic review of proctoring systems (conventional, online, automated, AI-based), analyzing applications, technologies, challenges, and research gaps across eight research questions covering features, challenges, countermeasures, and future trends. It synthesizes the literature on proctoring's evolution, categories, and associated challenges to guide future research.

What this means for practice

• Administrators. Pair any AI monitoring with accessible alternatives and a plan for test-taker anxiety, because the review documents false malpractice accusations when proctor and test-taker proficiency diverge, and infrastructure demands (webcam, microphone, reliable connection) that not all students can meet.
• Assessment professionals. Weigh surveillance against assessment redesign: the review's own conclusion is that remote and automated proctoring addresses the integrity needs of online assessment but is bounded by cost, infrastructure equity, and psychological burden.
• Administrators. Budget for the full infrastructure rather than the license: AI/ML proctoring that reads facial expressions and emotions through imaging raises both cost and hardware requirements, and real-time video storage plus CNN inference for gaze, head pose, and expression detection create latency and storage pressure.
• Instructors. Tell students in advance what the system records and why, since readiness, hardware and software requirements, and security and privacy concerns were the challenges students reported most in the review's case study, alongside the faculty role in the transition.
• Researchers. Treat gaze, head-pose, and activity recognition as unsolved in unconstrained settings: the review notes that deep-learning multi-user gaze estimation remains largely unexplored for lack of generic, publicly available datasets.

Limitations

• The search covers only IEEE, Science Direct, SCOPUS, and SCI plus Springer, EBSCO, ProQuest, and Taylor & Francis, restricted to 2008-2024 and to English-language peer-reviewed publications — an exclusion the authors note may drop research from non-English-speaking regions where online education is expanding fastest.
• Publication bias is acknowledged in the review itself: studies with statistically significant or positive results are more likely to be published, overrepresenting successful proctoring implementations and underreporting failures or ineffective systems.
• No primary data were collected. The review synthesizes other studies, and figures such as ~37.8% of college and ~41.8% of high-school students admitting to cheating come from the cited surveys rather than any measure the authors designed or administered.
• The authors state that AI-based proctoring systems are trained on Western behavioral norms and may not be trained on datasets from other regions, which limits the fairness claims of the systems reviewed.

Citation

Malhotra, M., & Chhabra, I. (2026). A comprehensive review of the changing landscape of academic dishonesty in automated proctoring in the era of artificial intelligence. Discover Education, 5, 236.