逻辑回归优化的DTW多点触摸身份认证算法

摘要

针对智能手机所面对的安全威胁问题,提出一种逻辑回归优化的动态时间规整身份认证算法.经手机触摸屏采集用户原始数据序列,使用平滑去噪和归一化处理后得出有效数据,并提取分类特征;提出一种改进距离计算方法并使用逻辑回归方法对识别特征进行特征级融合,通过优化后的动态时间规整算法来进行测试序列与参考序列的相似性度量;最后在身份认证阶段对单序列的认证结果采用与规则融合,得出最终认证结果.仿真结果表明,所提算法在用户身份识别率与识别稳定性上均优于传统动态时间规整算法,且检测虚假用户的性能优于SVDE和PSO-RBFN算法.%An algorithm of authentication,which is based on Dynamic Time Warping optimized by logistic regression,is proposed to defend the security threats attacking against smart phone.The original data sequences are collected from the phone touch screen completed by a user,and then the classification feature are extracted from the valid data obtained by smoothing and normalizing the original data.An improved distance calculation method and logistic regression is proposed to fusion the features,and the optimized Dynam ic Time Warping algorithm is proposed to measure the similarity between registered template sequences and test template sequences.The final authentication result is obtained by fusion the separate finger authentication results.Simulation results show that,compared with classical Dynamic Time Warping algorithm,the proposed algorithm has better performance on recognition rate and recognition stability,and detection of fake users outperforms SVDE and PSO-RBFN algorithm.