基于用户鼠标行为的身份认证方法

User Authentication Method Based on Mouse Dynamics

提出一种基于用户鼠标行为特征的身份认证方法,通过对用户的鼠标行为数据进行分析,归纳出两类鼠标行为特征,基于鼠标基本行为的一级特征,以及基于基本行为关系的二级特征,并采用极限学习机作为分类算法实现对用户身份的认证。通过实验表明所提出的二级特征能够有效降低认证方法的认假率,而极限学习机因其良好的泛化能力与优秀的学习速度是合适的认证算法。

This paper presents a user authentication method based on mouse dynamics as a behavioral biometric.Two kinds of features are proposed to characterize a user s unique pattern of the interaction with mouse device.One-level features are fine-granted metrics extracted for accurate characterization of five basic mouse actions and two-level features are defined based on the relations between the basic mouse actions.Extreme Learning Machine(ELM)is utilized for quick and efficient classification.The efficacy of the proposed features and verification method are evaluated by experiments based on a publicly available dataset.The experimental results demonstrate the high efficiency of ELM in verifying users’identities based on mouse dynamics.