中文摘要：

本文在Windows操作系统中研究并实现了三种截获用户鼠标行为数据的方法：消息钩子、WM_INPUT消息处理和过滤驱动，并从采样时钟分辨率、时间精度和位置信息等方面分析了不同方法所获取数据之间的区别．本文实验中同时使用这三种方法获取了多个用户的鼠标行为数据，从特征层面对不同数据截获方法带来的影响进行了对比分析，并利用神经网络分类器构建了身份认证模型，对使用不同方法下采集的鼠标行为数据进行用户身份认证的效果进行了实验研究．结果表明由三种数据截获方法采集到的数据都取得了较好的身份认证效果，其中使用消息钩子获取的数据进行认证的准确率最高，达到了94％以上．

英文摘要：

This paper presents a comparative analysis of three different mouse activity data interception techniques in the Windows operating system and studies their impacts on the performance of mouse dynamics based authentication. The differences between three interception techniques were analyzed from perspectives of sampling resolution, timer resolution and position information. A controlled environment for intercepting mouse activity data from multiple users using three interception techniques was designed. The effects of the different techniques were at first studied from the level of mouse activity features. Then an authentication model based on neural networks was established to classify mouse activity data, and the effect of the different interception techniques on the performance of authentication was evaluated. The results show the message hooking technique appears to be a better interception technique achieving an accuracy rate above 94%.