基于神经网络的人脸识别模型研究

Research on Face Recognition Model Based on Neural Network

生物识别技术常被用于网络安全领域以认证和授权访问为目的的身份识别过程中,用户提供的生物特征数据由数据安全系统采用的协议进行处理和转换,再与已提交和被认证的授权用户生物特征数据进行比对。比对的结果决定是否授予和授予该用户何等访问权限。在信息安全领域,生物特征识别方式是一个成熟和有效的安全验证机制,错误率低。近些年,网络安全面临着严峻形势,各信息安全认证系统中预存储的生物特征图像面临着被盗取并被入侵者滥用的风险。由此,该文提出一种不涉及修改存储用户生物特征图像的安全加固的存储系统,该系统将提交的生物特征图像生成用于认证和访问授权的无规律密码。为确保生物特征提取的准确性,该研究还引入深度学习模型将生物特征图像转换成二进制字符串形式存储。该研究通过实验计算得出一阶和二阶错误概率。实验结果表明,该文提出的加密系统不但实现可靠提取图像中生物特征的功能,还能保证生成的二进制字符串的高安全性和识别准确度。

Biometric technology is often used in the process of identity identification for the purpose of authentication and authorized access in the field of network security,where the biometric data provided by users are processed and converted by the protocols adopted by the data security system,and then compared with the biometric data of submitted and authenticated authorized users.The result of the comparison determines how access is granted and granted to the user.In the field of information security,biometric identification is a mature and effective security verification mechanism with low error rate.In recent years,network security is facing a severe situation,and the biometric images pre-stored in various information security authentication systems are facing the risk of being stolen and abused by intruders.Therefore,this paper proposes a secure storage system which does not involve modifying and storing user biometric images,which generates irregular passwords for authentication and access authorization from the submitted biometric images.In order to ensure the accuracy of biometric extraction,this study also introduces a deep learning model to convert biometric images into binary strings for storage.In this study,the first-order and second-order error probabilities are calculated by experiments.The experimental results show that the encryption system proposed in this paper not only realizes the function of reliably extracting biometric features from the image,but also ensures the high security and recognition accuracy of the generated binary string.