The SapiMouse database
The SapiMouse dataset was collected in 2020, and contains mouse dynamics data from 120 subjects (92 male and 28 female).
The subjects were from the Sapientia University with age between 18 and 53 years.
Data acquisition protocol
- Data collection was performed using a JavaScript web application running on the users' own computer.
- The data collection application is publicly available at https://mousedynamicsdatalogger.netlify.app .
- The participants used different browsers and mouses with a huge variety of DPI (dot-per-inch) settings, in some cases they used touch-pad.
- Movements were sampled by event-driven sampling technique, when users interact with their devices.
- The goal was accomplishing as much mouse operations as possible during the given time interval.
- Subjects were asked to complete two sessions, a three minutes session followed by a one minute session.
Session files
Each user has a dedicated folder with two files. The first one contains the data recorded during a session that lasted for 3 minutes, whereas the second file refers to the one-minute-long session, respectively.
The naming convention of the files is as follows: session_[date]_[1min|3min].csv, where date is the date of the data collection.
Each line of the file represents one mouse event and consists of the following features:
- timestamp: elapsed time since the start of the session
- button: describes the current condition of the mouse buttons
- state: contains information about the current state of the mouse
- x, y: the coordinates of the cursor on the screen
Download database
Publications
-
Margit Antal, Norbert Fejer, Krisztian Buza (2021), SapiMouse: Mouse Dynamics-based User Authentication Using Deep Feature Learning, May 19-21, 2021, [LINK]
-
M. Antal, K. Buza and N. Fejer (2021), "SapiAgent: A Bot Based on Deep Learning to Generate Human-Like Mouse Trajectories," in IEEE Access, vol. 9, pp. 124396-124408, 2021, doi: 10.1109/ACCESS.2021.3111098, IF: 3.367.
[Link]