Workshop on Statistical and Machine
Learning Techniques in Computer
Intrusion Detection
User Re-Authentication via Mouse Movements
Maja Pusara and Carla Brodley
Purdue University
Abstract:
We present an approach to user re-authentication based on the data collected
from the computer's mouse device. The underlying hypothesis is that one
can
successfully model user behavior based on the user-invoked mouse movements.
Our implemented system raises an alarm when the current behavior of user
X, deviates sufficiently from learned "normal" behavior of user
X. To learn a model of normal behavior we explore unsupervised learning
(learn a model for user X only) and supervised learning (learn to discriminate
among the behaviors of the k users of the computer). Our empirical results
show that although individuals that use the same set of applications can
be
differentiated via their mouse behavior, analyzing mouse movements alone
is
not sufficient for a stand-alone intrusion detection system. We conclude
that user re-authentication via mouse movements is one line of defense
in the hierarchical structure of an intrusion detection system.