Mouse Movement Authentication
for Multiple-Choice Tests

Mouse motion behavior is believed to be unique to an individual. The way a user moves a mouse in an application depends on hand shape and size, muscle control, and experience with the system. Mouse motion is ubiquitous for desktop computer users so there is much motivation to be able to identify and verify a user based on mouse movements.

Similar to keystroke biometric studies, two primary types of mouse movement studies have been conducted - those on a fixed sequence of mouse moves and those on arbitrary mouse input, and the arbitrary input can be structured to various degrees [see refs].

There are several applications that can utilize mouse movement biometric information, such as intrusion detection and enhancing/augmenting keystroke and other biometrics. For many biometrics the research interest is to determine the level of accuracy possible in various experimental situations as a function of the quantity of input data and the number of users. Recent project work focused on verifying the user's identity on various types of structured, application-specific mouse movement input [refs].


This project will focus on authenticating online test takers using mouse input. We will use previously collected mouse motion data from users taking exams on Moodle over several semesters. We currently have over 50 users who have taken 10 quizzes, each quiz containing 10 multiple-choice questions.

The project team will develop a machine learning biometric authentication system to authenticate test takers. A biometric authentication system provides a yes/no binary response -- yes you are the person you claim to be or no you are not. Machine learning systems generally have two processing steps -- converting the raw data input into a vector of features and then classifying the feature vector into a yes/no response. The features have been designed in the earlier project work but need to be coded, and the team could add additional features once the system is operational. For converting the feature vector into the yes/no response, various classification engines are available in Python, R language, Matlab, etc.

This is one of the more difficult projects but good work on this project will be strongly rewarded gradewise and can result in published conference papers and journal articles.

References (first is the most important)

  1. 2016 Fall Project Paper.
  2. Francis Buckley, Vito Barnes, Thomas Corum, Stephen Gelardi, Keith Rainsford, Phil Dressner, and John V. Monaco, Design of the Data Input Structure for a Mouse Movement Biometric System to Authenticate the Identity of Online Test Takers, Proc. Research Day, CSIS, Pace University, May 2015.
  3. Hedieh Zandikarimi, Frank Lin, Celia Carlos, Justin Correa, Phil Dressner, and Vinnie Monaco, Design of a Mouse Movement Biometric System to Verify the Identity of Students Taking Multiple-Choice Online Tests, Proc. Research Day, CSIS, Pace University, May 2014.
  4. Francisco Betances, Adam Pine, Gerald Thompson, Hedieh Zandikarimi, and Vinnie Monaco, Mouse Biometric Authentication, Proc. Research Day, CSIS, Pace University, May 2014.
  5. Pedro Xavier de Oliveira, Venugopala Channarayappa, Eamonn O'Donnel, Bappaditya Sinha, Aswinkumar Vadakkencherry, Tushar Londhe, Umesh Gatkal, Ned Bakelman, John V. Monaco, and Charles C. Tappert, Mouse Movement Biometric System, Proc. Research Day, CSIS, Pace University, 2013. slides
  6. Chao Shen, et al., User Authentication Through Mouse Dynamics, IEEE Trans. Info. Forensics and Security, 8-1, Jan 2013.
  7. Allen Newell, Section on Fitts' Law, Unified Theories of Cognition (The William James Lectures), Harvard University Press, 1994.
  8. Nkem Ajufor, Antony Amalraj, Rafael Diaz, Mohammed Islam, Michael Lampe, Refinement of a Mouse Movement Biometric System, Proc. Research Day, CSIS, Pace University, 2007.
  9. Nan Zheng, Aaron Paloski, and Haining Wang, An Efficient User Verification System via Mouse Movements, CCS’11, Chicago, October 2011.
  10. Maja Pusara and Carla E. Brodley, User Re-Authentication via Mouse Movements, VizSEC/DMSEC'04, Washington DC, 2004.