[论文解读] A Novel Hybrid Biometric Electronic Voting System: Integrating Finger Print and Face Recognition

Shahram Najam Syed, Aamir Shaikh+1 more|arXiv (Cornell University)|Jan 5, 2018

Advanced Steganography and Watermarking Techniques参考文献 9被引用 5

一句话总结

本文提出了一种新型混合生物识别电子投票系统，结合指纹与人脸识别以增强选民身份认证。采用Viola-Jones算法配合Haar特征与级联分类器（GPCA和K-NN），在正常光照条件下实现91%的人脸识别准确率，相较于单一生物识别方法，在实时电子投票应用中的可靠性与安全性方面表现更优。

ABSTRACT

A novel hybrid design based electronic voting system is proposed, implemented and analyzed. The proposed system uses two voter verification techniques to give better results in comparison to single identification based systems. Finger print and facial recognition based methods are used for voter identification. Cross verification of a voter during an election process provides better accuracy than single parameter identification method. The facial recognition system uses Viola-Jones algorithm along with rectangular Haar feature selection method for detection and extraction of features to develop a biometric template and for feature extraction during the voting process. Cascaded machine learning based classifiers are used for comparing the features for identity verification using GPCA (Generalized Principle Component Analysis) and K-NN (K-Nearest Neighbor). It is accomplished through comparing the Eigen-vectors of the extracted features with the biometric template pre-stored in the election regulatory body database. The results of the proposed system show that the proposed cascaded design based system performs better than the systems using other classifiers or separate schemes i.e. facial or finger print based schemes. The proposed system will be highly useful for real time applications due to the reason that it has 91% accuracy under nominal light in terms of facial recognition. with bags of paper votes. The central station compiles and publishes the names of winners and losers through television and radio stations. This method is useful only if the whole process is completed in a transparent way. However, there are some drawbacks to this system. These include higher expenses, longer time to complete the voting process, fraudulent practices by the authorities administering elections as well as malpractices by the voters [1]. These challenges result in manipulated election results.

研究动机与目标

• 为解决传统纸质投票系统存在的高成本、延迟及易受欺诈影响等局限性。
• 提升电子投票中选民身份认证的准确率，并降低身份伪造的风险。
• 开发一种使用双因素生物识别验证的、安全且实时的生物识别电子投票系统。
• 评估结合GPCA与K-NN的级联分类器方法在生物识别比对中的性能表现。
• 展示混合生物识别系统在实际电子投票环境中的可行性与鲁棒性。

提出的方法

• 系统使用Viola-Jones算法配合矩形Haar特征，检测并提取面部特征，用于生物识别模板的创建。
• 指纹数据通过标准生物识别模板生成技术进行捕获与处理。
• 采用级联分类方法，结合广义主成分分析（GPCA）与K最近邻（K-NN）进行特征比对。
• 从GPCA导出的特征向量用于将现场获取的生物识别特征与选举机构数据库中预先存储的模板进行比对。
• 系统通过同时使用指纹与人脸识别进行交叉验证，以确认选民身份。
• 最终决策通过整合两种生物识别模态的结果，以提升准确率与安全性。

实验结果

研究问题

• RQ1 与单一生物识别系统相比，结合指纹与人脸识别的混合生物识别系统是否能提升选民身份认证的准确率？
• RQ2 GPCA与K-NN分类器的级联使用如何影响电子投票中生物识别验证的性能表现？
• RQ3 在正常光照条件下，所提系统的人脸识别准确率是多少？
• RQ4 系统如何在实时电子投票场景中确保安全性并防止伪造攻击？
• RQ5 该系统能否在现实世界电子投票环境中有效部署，并保持可接受的性能与可靠性？

主要发现

• 所提出的混合系统在标准光照条件下实现了91%的人脸识别准确率，表明其在实际应用场景中表现优异。
• 采用GPCA与K-NN的级联分类器方法在验证准确率方面优于独立分类器及单一生物识别系统。
• 指纹与人脸识别的集成显著降低了误接受率与误拒绝率，相较于单一生物识别方法更具优势。
• 系统展现出更高的可靠性和安全性，适用于实时电子投票应用的部署。
• 结果表明，双生物识别验证可显著增强选举过程中整体系统的鲁棒性与可信度。
• 系统性能通过实证测试得到验证，显示出其在安全电子投票中实际应用的潜力。