Privacy-Preserving On-Screen Activity Tracking and Classification in E-Learning using AI

Abstract

The blistering development of online education has changed considerably the manner in which education, learning as well as scholarly analysis is given in academic institutions, instructional centers, and in professional education. Internet technologies have provided the flexibility in accessing learning materials regardless of geographical limits and thus learning institutions have been able to touch the greater and more varied populations of learners. Nevertheless, the task of making online assessment secure and trustworthy is a pressing challenge despite the fact that the process of content delivery has been developed at a remarkable pace. Invigilators monitor student behaviour directly in physical examination halls, confirm identity, limit unauthorised materials and compliance with examination regulations. It is much more challenging to provide the identical level of control within remote settings.

Somewhat towards this end, various scientists and organizations have been undertaking the study of learning intelligent proctoring systems which monitor student behaviour automatically in online exams. These are the systems that usually use the help of such tools as webcams, microphones, screen-capture functions, and behavioural analytics to detect suspicious behaviour like impersonation, collaborating with third parties, unusual gazes movement, multiple-people presence, using a hidden mobile phone, and constantly switching applications. Recent developments in artificial intelligence, particularly computer vision and machine learning, have enhanced the ability of such systems to scan facial expressions, head pose, attention patterns and environmental cues in real time.

Although these have some advantages, issues about privacy, lack of fairness and bias in algorithmic choices, over-surveillance and false charges have hindered universal adoption. Thus, the new generation of online monitoring tools needs to strike a balance between integrity in examinations and responsible data management. In this review, a detailed discussion will be made on privacy preserving activities tracked on-screen in e-learning assessment. It examines significant technologies, the current constraints, architectural design imperatives, ethical issues, and future research potential. An approach is also outlined that uses a privacy-conscious framework, where sensitive processing of data is carried out locally, but significant risk indicators and little encrypted evidence are sent to reviewers. These solutions are able to enhance trust, scalability and regulatory compliance of digital education e-mails.