基于轻量神经网络的指静脉密钥生成方法
Finger Vein Key Generation Method Based on Lightweight Neural Network
周洋 1王明文1
作者信息
- 1. 西南交通大学 数学学院,四川 成都 611756
- 折叠
摘要
针对生物特征密钥生成中容易泄露模板信息、精度性能不高和过于复杂难以运用等问题,提出一种基于轻量神经网络的指静脉密钥生成方法.以反向残差瓶颈结构为核心提出轻量神经网络,结合标签平滑交叉熵对指静脉图像训练处理.在密钥生成模块中提出随机选择模块,快速生成新的密钥.最后采用纠错技术与安全散列算法,解决网络提取特征的不稳定性,增强密钥生成方法的安全性.该方法在三个公开数据库中得到了验证,提出的方法生成512位密钥的误识率为0.843%～1.469%、拒真率为0.651%～1.524%,并且密钥生成耗时不超过0.3 s,获得了比其他方法更优越的性能.安全分析表明,提出的模型可以有效抵御信息泄露、交叉匹配和其他攻击.理论分析和实验结果表明该方法具有泛化能力强,生成密钥精度高、生成时间短、安全性高等性质.
Abstract
Aiming at the problems of easy disclosure of template information,low accuracy performance,complexity,and difficulties in generating biometric key,a finger vein key generation method based on lightweight neural network was proposed.Taking the in-verted residual bottleneck structure as the core,a lightweight neural network was proposed,the finger vein image training processing was combined with label smoothing cross-entropy.In the key generation module,a random selection module was proposed to gener-ate new keys quickly.Finally,error correction technology and security hash algorithm were used to solve the instability of network feature extraction and enhance the security of key generation method.This method has been verified in three public databases,and the false acceptance rate of the proposed method to generate a 512-bit key is 0.843%-1.469%,the false rejection rate is 0.651%-1.524%,and the key generation time does not exceed 0.3 s,which shows superior performance than other methods.Security analy-sis shows that the proposed model can effectively resist information leakage,cross-matching and other attacks.Theoretical analysis and experimental results show that the proposed method has the properties of strong generalization ability,high key generation accu-racy,short generation time and high security.
关键词
生物特征密钥/轻量神经网络/反向残差瓶颈结构/安全性分析Key words
biometric keys/lightweight neural networks/inverted residual bottleneck structure/security analysis引用本文复制引用
基金项目
国家自然科学基金(62106206)
四川省科技计划项目(2020YFG0045)
出版年
2024
国家科技期刊平台
NSTL
万方数据