基于集成学习和双并行自适应机制的击键动力学认证方法  

Keystroke dynamics authentication method based on ensemble learning and dual parallel adaptive mechanism

作　　者：崔立军 于宝华[1] 荣江 CUI Lijun;YU Baohua;RONG Jiang(College of Information Science and Technology,Shihezi University,Shihezi,Xinjiang 832003,China;Network Information Center,Xinjiang University of Political Science and Law,Tumxuk,Xinjiang 844000,China)

机构地区：[1]石河子大学信息科学与技术学院,新疆石河子832003 [2]新疆政法学院网络信息中心,新疆图木舒克844000

出　　处：《石河子大学学报（自然科学版）》2024年第4期495-504,共10页Journal of Shihezi University(Natural Science)

基　　金：新疆生产建设兵团财政科技计划项目(2020DB005,2021AB023)。

摘　　要：身份认证是指在计算机系统中确认操作者身份的过程,击键动力学作为一种成本低廉、难以模仿的身份认证方式得到许多学者的广泛关注。然而,现有的方法往往存在误判率和漏判率偏高、泛化能力差等弊端。针对以上问题,本文提出一种将集成学习和自适应更新机制结合的方式,在提高模型分类性能的同时适应新数据中的特征变化。通过使用公开的CMU数据集和通用的评估指标(EER)将本文的方法与其他先进的技术进行比较,实验表明本文所提出的二次集成学习方法性能优异,使用双并行自适应更新机制后表现出可靠的泛化能力,在CMU数据集上得到了3.22%的EER,模型性能优于相同实验条件下的同类研究。Identity authentication refers to the process of confirming the identity of an operator in a computer system.Keystroke dynamics,as a low-cost and difficult to imitate method of identity authentication,has received widespread attention from many scholars.However,existing methods often have drawbacks such as high false positive and false negative rates,and poor generalization ability.In response to the above issues,this article proposes a method that combines ensemble learning and adaptive update mechanism to improve the classification performance of the model while adapting to feature changes in new data.By comparing our method with other advanced technologies using publicly available CMU datasets and universal evaluation metrics(EER),experiments show that our proposed quadratic ensemble learning method has excellent performance.After using a dual parallel adaptive update mechanism,it exhibits reliable generalization ability,achieving an EER of 3.22%on the CMU dataset.The model performance is better than similar studies under the same experimental conditions.

关 键 词：身份认证 击键动力学 集成学习 自适应更新 

分 类 号：TP393.0[自动化与计算机技术—计算机应用技术]