Abstract: A new user authentication approach based on users’ keystroke patterns using manifold learning is proposed. The proposed approach utilizes Geodesic distance to denote the difference between sample vectors and then uses a new nonlinear dimensionality reduction algorithm: isometric mapping (ISOMAP) to find intrinsic geometry structure hiding in users’ keystroke patterns space. The performance of this approach is evaluated using 1500 keystroke sequences，the experimental results show the superiority of this approach in terms of false reject rate(FRR) and false accept rate(FAR) compared with some recent existing methods, the FRR and FAR of the proposed approach is only 1.65% and 0% respectively.

摘要： 本文基于等距映射（ISOMAP）非线性降维算法, 提出了一种新的基于用户击键特征的用户身份认证算法, 该算法用测地距离代替传统的欧氏距离, 作为样本向量之间的距离度量,在用户击键特征向量空间中挖掘嵌入的低维黎曼流形，进行用户识别。用采集到的1500个击键模式数据进行实验测试,结果表明,该文的算法性能优于现有的同类算法,其错误拒绝率（FRR）和错误通过率（FAR）分别是1.65%和0%，低于现有的同类算法。