Keystroke and Mouse User Behavior Data

Our background analysis confirms that there are no publicly available datasets focused on online assessment fraud detection in online education platforms. Additionally, there is a lack of research
studies that examine the evolution of user behavior and the underlying reasons for these changes. Factors contributing to this behavioral evolution have also been largely unexplored, with no studies found that actively attempt to analyze this information.

We developed an online assessment game using HTML, JavaScript, and CSS, which includes two distinct activity types: one designed to collect keystroke data and the other to capture mouse behavior data. The keystroke activity comprises four tasks—two question-answering tasks and two paragraph-copying tasks—each performed once by the user. The mouse behavior activity also consists of four tasks: Multiple Choice Questions (MCQ), Drag-and-Drop, Click-the-Target, and Matching tasks. Users complete the MCQ, Drag-and-Drop, and Matching tasks 10 times each.

It is important to note that data collection spans over a year, with user sessions spaced approximately one month apart. This interval is intentional, allowing us to track and analyze variations in user behavior over time.

Software/equipment used to create/collect the data: We will be developing a online assessment game using HTML, CSS, JavaScript, and MongoDB to collect and store information. This online assessment game will be website which can loaded on participants system browsers and does not need any specific installation procedure. Participants will only need to register and then login to play the game

Software/equipment used to manipulate/analyse the data: The main software will either manipulated or analyzed using personal systems (Asus TUF Laptop). Anaconda software especially jupyter notebooks, spyder will used for programming and training AI algorithms. Alternatively, Google colabs will also be used for conducting experimentation.