基于三分类的击键序列身份认证

针对基于统计学用户击键模式识别算法识别率较低的不足，提出了一种统计学三分类主机用户身份认证算法。该方法通过对当前注册用户的击键特征与由训练样本得到的标准击键特征进行比较，将当前注册用户划分为合法用户类、怀疑类与入侵类三类,对怀疑类采用二次识别机制。 采用动态判别域值，引入了与系统安全性和友好性相关的可控参量k，由系统管理员根据实际确定。并对该算法性能进行了理论分析与实验测试，结果表明该算法在保持贝叶斯统计算法需要训练样本集规模较小、算法收敛速度快优点的基础上,识别精度高于贝叶斯统计算法，错误拒绝率(FRR)和错误通过率(FAR)分别为1.6%和1.5%。

Statistical three-class user authentication approach based on user's keystroke patterns

In order to overcome the deficiency of low veracity of the current user authentication approach based on statistics,a new user authentication approach based on keystroke sequences was proposed.The proposed approach classified the current registered users into three groups: intrusive,suspicious and normal according to the dispersion between the user's keystroke pattern and the standard model determined by training samples and then utilized a re-recognition process to identify the user belonging to the suspicious group.The parameter k related to the security and friendliness level was introduced,which can be dynamically determined by supervisors of hosts.The performance of the proposed approach was evaluated by experiments and theoretical analyses.The results show the superiority of the proposed approach in terms of the False Rejection Rate(FRR) and False Acceptance Rate(FAR) in comparison with current Bayesian method,and the FRR and FAR of the proposed approach is only 1.6% and 1.5% respectively.