基于顽健线性判别分析的击键特征识别方法Keystrokefeatures.PDFVIP

第38 卷第Z2 期 通 信 学 报 Vol.38 No.Z2 2017 年11 月 Journal on Communications November 2017 doi:10.11959/j.issn.1000-436x.2017272 基于顽健线性判别分析的击键特征识别方法 1,2 1,2 沈伟国 ，王巍 （1. 通信信息控制和安全技术重点实验室，浙江 嘉兴 314033; 2. 中国电子科技集团公司第三十六研究所，浙江 嘉兴 314033 ） 摘 要：研究了用户认证过程中的键盘击键序列特征提取和分类问题，提出一种基于顽健线性判别分析的击键特 征识别方法。首先，最大化击键序列集不同类间的离散度，同时最小化序列集同类之间的离散度，保持击键序列 样本的最佳判别特性。其次，最小化近邻击键序列样本间的相似性离散度，保持序列样本的区域相似性。最后， 基于上述原则，对击键序列特征样本进行特征提取，并采用最近邻分类准则进行判决输出。通过与其他方法的实 验对比，验证了该方法的有效性。 关键词：身份认证；击键序列识别；特征提取；顽健线性判别分析 中图分类号：TP391.4 文献标识码：A Keystroke features recognition based on stable linear discriminant analysis SHEN Wei-guo1,2, WANG Wei1,2 (1. Science and Technology on Communication Information Security Control Laboratory, Jiaxing 314033, China; 2. No.36 Research Institute of CETC, Jiaxing 314033, China) Abstract: A novel keystroke features recognition method based on stable linear discriminant Analysis (SLDA) was put forward. First of all, it maximum the dispersion between different sequences, while minimizing the dispersion between the same sequence set, maintain the best discriminant characteristics of the keystroke sequences. Secondly, the local si- milarity graph between keystroke sequences is constructed, minimizing the dispersion of the local similarity sequences, to keep the local similarity of keystroke sequences. Finally, based on the principles above, the feature of keystroke se- quences are extracted, and the nearest neighbor classificati