融合人脸特征与密码算法的身份认证系统

利用人脸识别技术、活体检测技术，结合国密算法，设计了一个基于模糊承诺算法的身份认证系统。提出一种多值量化方法来提高数据相似性，并采用多样本组的特征数据清洗方式提高数据稳定性，以解决常见的生物特征认证中误识率和拒识率较高且难以平衡的问题。在注册阶段，采集多组128维人脸特征数据，清洗后获得平均值，基于一个阈值区间将每个维度的数据量化为4 b二进制数。将量化完成后的数据作为加密密钥，以BCH编码为纠错码，使用模糊承诺算法将认证服务器产生的秘密密钥加密存储在客户端。在认证阶段，实时采集的人脸特征数据经过量化后，利用BCH纠错提取出秘密密钥，将秘密密钥作为协商密钥，基于传统身份认证协议实现客户端与认证服务器之间的认证过程。通过实验证实，采用上述方法实现的身份认证系统可将误识率降低至0%,拒识率降低到1%以内。

An identity authentication system integrating facial features and cipher algorithm

An identity authentication system based on fuzzy commitment algorithm is designed by using face recognition technology,living body detection technology and national secret algorithm.In order to solve the problems of high false accept rate and high false reject rate and hard to balance in common biometric authentication,a multi-value quantification method is proposed,and the feature data cleaning method of multi-sample groups is adopted to improve data stability.During the registration phase,multiple sets of 128-dimensional face feature data are collected,and the average value is obtained after data cleaning.Based on a threshold interval,the data of each dimension is quantized into a 4 b binary number.By using the quantized data as the encryption key,and the BCH code as the error correction code,the secret key generated by the authentication server is encrypted and saved on client by the fuzzy commitment algorithm.During the authentication stage,after the face feature data collected in real time is quantified,the secret key is extracted by BCH error correction,which is used as the negotiation key.And the authentication process between the client and the authentication server is realized according to the traditional identity authentication protocol.Experiments confirm that the identity authentication system implemented by above method reduces the false recognition rate to 0% and the rejection rate to less than 1%.