FHE~* KDFRS:全同态加密相容的核基人脸识别系统
|
摘要
生物特征识别是一种有着特征唯一、不易复制等良好特性的个人身份鉴定与识别技术.但在识别过程中,个人信息通过公开信道传输或网络服务器存储时,有可能会受到第三方的截获和修改,或通信双方提供虚假信息进行相互欺骗.可通过引入全同态加密协议以保护数据与分类器.此类方案设计主要存在2方面问题:一方面是只支持"加乘"运算的全同态加密算法与识别算法的运算相容性问题;另一方面是由于加密算法的约束导致识别率与运行效率的降低.以C/S模型为基础,采用了Gabor小波和核主成分分析法,利用数据的非线性信息和高阶统计特性以提高识别率;并设计了通信协议,使用了多项式核和改进后的DGHV加密方案,以解决相容性问题.原型实现的实验数据表明,该方案在承接源自全同态加密的隐私保护的前提下,有着较高的识别率与运行效率,其累积匹配率为91.9%,最高识别率为97.62%,最大识别时间花销约为1 s.
Biometrics is a kind of personal identification and recognition technology where the personalized face data are unique and not easily replicated.But when personal information transmitted through public network channel or stored through a network sever,it is likely to be intercepted and modified by the third party or deceived by other side of communication in the process of identification.To ensure data security,data and classifiers are encrypted by introducing cryptographic protocols.There are two main challenges in solving the privacy-preserving design: one is the compatibility problem between the cryptographic protocols and the kernel based recognition algorithms,the other is the low recognition rate and the low recognition efficiency caused by the constraint of the selected encryption algorithms with only support for " multiplication and addition".To keep the recognition rate,the Gabor wavelets and K-PCAs are used for feature extractions which try to retain data nonlinearity and higher-order statistics as far as possible based on the C/S model; To meet the demand for operation compatibility,communication protocols are designed by utilizing polynomial kernel and the improved DGHV homomorphic encryption scheme.Experiments of prototype implementation show that the proposed scheme,which based on the premise of privacy protection from FHE,has the appropriate time and acceptable recognition rate for cumulative matching rate was 91.9%,the highest was 97.62%,and a maximum recognition time was about 1 s.
Biometrics is a kind of personal identification and recognition technology where the personalized face data are unique and not easily replicated.But when personal information transmitted through public network channel or stored through a network sever,it is likely to be intercepted and modified by the third party or deceived by other side of communication in the process of identification.To ensure data security,data and classifiers are encrypted by introducing cryptographic protocols.There are two main challenges in solving the privacy-preserving design: one is the compatibility problem between the cryptographic protocols and the kernel based recognition algorithms,the other is the low recognition rate and the low recognition efficiency caused by the constraint of the selected encryption algorithms with only support for " multiplication and addition".To keep the recognition rate,the Gabor wavelets and K-PCAs are used for feature extractions which try to retain data nonlinearity and higher-order statistics as far as possible based on the C/S model; To meet the demand for operation compatibility,communication protocols are designed by utilizing polynomial kernel and the improved DGHV homomorphic encryption scheme.Experiments of prototype implementation show that the proposed scheme,which based on the premise of privacy protection from FHE,has the appropriate time and acceptable recognition rate for cumulative matching rate was 91.9%,the highest was 97.62%,and a maximum recognition time was about 1 s.
引文
[1]陆正福,王欢.FHE-相容的分布式人脸识别方案设计与分析[J].计算机工程与设计,2016(6):1 428-1 434.LU Z F,WANG H.Design and analysis of FHE-Compatible distributed face recognition scheme[J]. Computer Engineering and Design,2016(6):1 428-1 434.
[2] ERKIN Z,FRANZ M,GUAJARDO J,et al.Privacy-preserving face recognition[C]//Privacy Enhancing Technologies,International Symposium,Seattle,Wa,USA.Proceedings DBLP,2009:235-253.
[3] SADEGHI A R,SCHNEIDER T,WEHRENBERG I.Efficient privacy-preserving face recognition[C]//International Conference on Information Security and Cryptology,Springer-Verlag,2009:229-244.
[4] TURK M,PENTLAND A.Eigenfaces for recognition[J].Journal of Cognitive Neuroscience,1991,3(1):71-86.
[5] UPMANYU M,NAMBOODIRI A M,SRINATHAN K,et al.Blind authentication:a secure crypto-biometric cerification protocol[J]. Proc IEEE Transactions on Information Forensics and Security,2010,5(2):255-268.
[6] OSADCHY M,PINKAS B,JARROUS A,et al.SCi FI-A system for secure face identification[C]//IEEE Security and Privacy,2010:239-254.
[7] HSU C Y,LU C S,PEI S C.Image feature extraction in encrypted domain with privary-preserving SIFT[J].IEEE Transactions on Image Processing,2012,21(11):4 593-4 607.
[8] TRONCOSO-PASTORIZA J R,GONZLEZ-JIMNEZ D,PREZ-GONZLEZ F. Fully private noninteractive face verification[J]. IEEE Transactions on Information Forensics&Security,2013,8(7):1 101-1 114.
[9]李航.统计学习方法[M].北京:清华大学出版社,2013.LI H.Statistical learning method[M]. Beijing:Tsinghua University Press,2013.
[10] GUMUS E,KILIC N,SERTBAS A,et al.Evaluation of face recognition techniques using PCA,wavelets and SVM[J]. Expert Systems with Applications,2010,37(9):6 404-6 408.
[11]叶超.基于Gabor小波和SVM的人脸识别算法研究[D].太原:中北大学,2014.YE C. The optimzation algorithm of face recognition based on Gabor wavelet and SVM[M]. Taiyuan:North Universty of China,2014.
[12] YANG M H. Kernel eigenfaces vs. Kernel fisherfaces:face Recognition using kernel methods[C]//Proceedings IEEE International Conference on Automatic Face and Gesture Recognition,2002:215-220.
[13]李云峰.基于Gabor小波变换的人脸识别[D].大连:大连理工大学,2006.LI Y F.Gabor wavelet transform based face recognition[D].Dalian:Dalian University of Technology,2016.
[14] DIJK M V,GENTRY C,HALEVI S,et al.Fully homomorphic encryption over the integers[D].Springer Berlin Heidelberg,2010.
[15] LINDELL Y,PINKAS B.A proof of Yao's protocol for secure two-party computation[J]. Electronic Colloquium on Computational Complexity,2004:2004.
[16] CORON J,MANDAL A,NACCACHE D,et al.Fully homomorphic encryption over the integers with shorter public keys[C]//Conference on Advances in Cryptology,Springer-Verlag,2015:487-504.
[17] HUANG J,YUEN P C,CHEN W S,et al.Choosing parameters of kernel subspace LDA for recognition of face images under pose and illumination variations[J].IEEE Transactions on Systems Man&Cybernetics Part B Cybernetics,2007,37(4):847-862.
[2] ERKIN Z,FRANZ M,GUAJARDO J,et al.Privacy-preserving face recognition[C]//Privacy Enhancing Technologies,International Symposium,Seattle,Wa,USA.Proceedings DBLP,2009:235-253.
[3] SADEGHI A R,SCHNEIDER T,WEHRENBERG I.Efficient privacy-preserving face recognition[C]//International Conference on Information Security and Cryptology,Springer-Verlag,2009:229-244.
[4] TURK M,PENTLAND A.Eigenfaces for recognition[J].Journal of Cognitive Neuroscience,1991,3(1):71-86.
[5] UPMANYU M,NAMBOODIRI A M,SRINATHAN K,et al.Blind authentication:a secure crypto-biometric cerification protocol[J]. Proc IEEE Transactions on Information Forensics and Security,2010,5(2):255-268.
[6] OSADCHY M,PINKAS B,JARROUS A,et al.SCi FI-A system for secure face identification[C]//IEEE Security and Privacy,2010:239-254.
[7] HSU C Y,LU C S,PEI S C.Image feature extraction in encrypted domain with privary-preserving SIFT[J].IEEE Transactions on Image Processing,2012,21(11):4 593-4 607.
[8] TRONCOSO-PASTORIZA J R,GONZLEZ-JIMNEZ D,PREZ-GONZLEZ F. Fully private noninteractive face verification[J]. IEEE Transactions on Information Forensics&Security,2013,8(7):1 101-1 114.
[9]李航.统计学习方法[M].北京:清华大学出版社,2013.LI H.Statistical learning method[M]. Beijing:Tsinghua University Press,2013.
[10] GUMUS E,KILIC N,SERTBAS A,et al.Evaluation of face recognition techniques using PCA,wavelets and SVM[J]. Expert Systems with Applications,2010,37(9):6 404-6 408.
[11]叶超.基于Gabor小波和SVM的人脸识别算法研究[D].太原:中北大学,2014.YE C. The optimzation algorithm of face recognition based on Gabor wavelet and SVM[M]. Taiyuan:North Universty of China,2014.
[12] YANG M H. Kernel eigenfaces vs. Kernel fisherfaces:face Recognition using kernel methods[C]//Proceedings IEEE International Conference on Automatic Face and Gesture Recognition,2002:215-220.
[13]李云峰.基于Gabor小波变换的人脸识别[D].大连:大连理工大学,2006.LI Y F.Gabor wavelet transform based face recognition[D].Dalian:Dalian University of Technology,2016.
[14] DIJK M V,GENTRY C,HALEVI S,et al.Fully homomorphic encryption over the integers[D].Springer Berlin Heidelberg,2010.
[15] LINDELL Y,PINKAS B.A proof of Yao's protocol for secure two-party computation[J]. Electronic Colloquium on Computational Complexity,2004:2004.
[16] CORON J,MANDAL A,NACCACHE D,et al.Fully homomorphic encryption over the integers with shorter public keys[C]//Conference on Advances in Cryptology,Springer-Verlag,2015:487-504.
[17] HUANG J,YUEN P C,CHEN W S,et al.Choosing parameters of kernel subspace LDA for recognition of face images under pose and illumination variations[J].IEEE Transactions on Systems Man&Cybernetics Part B Cybernetics,2007,37(4):847-862.