Poster: can it be more practical?: improving mouse dynamics biometric performance

Mouse dynamics is the process of verifying the identity of computer users on the basis of their mouse operating characteristics, which are derived from the movement and click events. Some researchers have explored this domain and reported encouraging results, but few focused on applicability in a realistic setting. Specifically, many of the existing approaches require an impractically long verification time to achieve a reasonable accuracy. In this work, we investigate the mouse dynamics of 26 subjects under a tightly-controlled environment. Using procedural features such as speed and acceleration curves to more accurately characterize mouse activity, and adopting distance metrics to overcome the within-class variability, we achieved a promising performance with a false-acceptance rate of 8.87%, a false-rejection rate of 7.16%, and an average verification time of 11.8 seconds. We find that while this level of accuracy comes close to meeting the requirements of identity verification, a tradeoff must be made between security and user acceptability. We also suggest opportunities for further investigation through additional, controlled experimental environments.