论文部分内容阅读
由于计算机用户对键盘的熟悉程度、击键习惯等不尽相同,每个用户都具有自己独特的击键生物特征.对于某个用户来说,其击键特征为正常类,其他所有用户为异常类,这可以利用模式识别中的单类分类器来解决.本文设计基于支持向量数据描述(SVDD)的击键生物特征身份认证系统模型.将该方法与 BP、RBF 和 SOM 方法进行对比,证实 SVDD 具有较好的识别效果,它可将非法用户误接受率从28.9%降低到0.28%.最后给出一个嵌入Windows 用户登录中的口令+击键特征身份认证的实现技术.
Due to computer users familiarity with the keyboard, keystrokes and other habits vary, each user has its own unique keystroke biometrics.For a user, the keystroke features for the normal class, all other users are abnormal Class, which can be solved by using a single classifier in pattern recognition.This paper designs a keystroke biometric identity authentication system model based on Support Vector Data Description (SVDD), which is compared with BP, RBF and SOM methods to confirm SVDD has a good recognition effect, which can reduce the false acceptance rate of unauthorized users from 28.9% to 0.28% .Finally, an implementation technique of password + keystroke signature authentication embedded in Windows user login is given.