CN115278673A - Lightweight biometric authentication method and system based on joint biometrics - Google Patents

Abstract

In the method and the system for lightweight biometric authentication based on joint biometric identification, a trusted center generates a series of keys; the extractor constructs a d-vitamin feature registration vector and a biological feature template by confusing the biological feature template, and encrypts the feature template by using a public key of a homomorphic encryption algorithm; the extractor expands the registration vector, encrypts and sends the registration vector to a calculation server by using a registration key, and encrypts and sends an index to a database by the calculation server; the extractor expands the characteristic vector, encrypts the expanded vector and the biological characteristic template and sends the encrypted expanded vector and the encrypted biological characteristic template to the computing server; the database transforms the received index and the authentication query, calculates the similarity between each registration template and the authentication query, and sends the candidate template set to a calculation server to calculate the Euclidean distance; three update operations are supported: add, delete, and modify; the invention meets the confidentiality, renewability, revocable, irreversibility and non-connectability of biological identification, and realizes the balance of low-cost authentication and high-safety requirements.

Description

Lightweight biometric authentication method and system based on joint biometrics

technical field

The invention belongs to the technical field of biometric identification, and in particular relates to a lightweight biometric authentication method and system based on joint biometric identification.

Background technique

In recent years, the popularity of smart mobile devices has continuously improved people's quality of life, and the market size of mobile devices is also expanding. In addition, smart watches, tablet computers and other mobile devices are also driving the expansion of the mobile market. While mobile devices bring convenience to people, they also pose a threat to users' privacy and security. As users store personal sensitive information (such as bank account and image data) on smart mobile devices, the leakage of personal privacy has also become the focus of researchers.

Most existing smart mobile devices use knowledge-based identity authentication mechanisms to ensure their own security and data privacy (such as PIN code-based, pattern-based password authentication). However, most users tend to set simple and weak passwords that are easy to remember. This knowledge-based authentication is vulnerable to snooping and dictionary attacks, whereby attackers can gain access to sensitive personal information stored on the device. Using its uniqueness, versatility, stability and availability, biometric technology has promoted the continuous development of biometric authentication, making biometric authentication more convenient and accurate. It also overcomes the vulnerability of password settings in knowledge-based authentication.

Through the above analysis, the problems and defects of the existing technology are: most of the existing biometric authentication methods are based on a single biometric feature, the accuracy and stability are not high, and they cannot be applied to different application backgrounds; in addition, the existing biometric authentication methods The security of the method is not high. Once the biometric feature is stolen, damaged or forged, the identity authentication based on the biometric feature may be threatened by artificial synthesis, replay, and spoofing attacks.

The difficulties in solving the above problems and defects are as follows: (1) The computing power and storage capacity of smart mobile devices are limited, so it is necessary to design a lightweight biometric authentication method. (2) Most of the existing biometric authentication methods are based on a single biometric feature, the accuracy and stability are not high, and they cannot be applied to different application backgrounds, so the designed method needs to be able to integrate these biometric features to achieve various biometric features Comprehensive application of information. (3) Due to the uniqueness of biometrics, once the biometrics are stolen, damaged or forged, identity authentication based on biometrics may be threatened by artificial synthesis, replay, and spoofing attacks. Therefore, the designed method needs to be able to reconstruct the user biometric template after the biometric is stolen or damaged.

Contents of the invention

In order to overcome the shortcomings of the above-mentioned prior art, the purpose of the present invention is to provide a lightweight biometric authentication method and system based on joint biometrics, which is suitable for the application environment of smart mobile devices at the present stage, and combines knowledge-based authentication and authentication-based The authentication of multiple biometrics can overcome the low security of password authentication only, and the application of the cancelable template module can prevent the irrecoverability of the biometric template after being stolen or damaged, which has the advantages of high security and low overhead.

In order to achieve the above object, the technical scheme that the present invention takes is:

A lightweight biometric authentication method based on joint biometrics, comprising the following steps:

私钥

对称加密密钥

和认证密钥

为注册用户R_i生成索引构建密钥

S101: The trusted center TA generates a series of keys, and generates a public key for the authenticated user U _i

private key

Symmetric encryption key

and authentication key

Generate an index build key for registered user R _i

S102: The extractor constructs a d-dimensional biometric registration vector v _R and a biometric template T _i by confusing the biometric template v _B , and encrypts the biometric template T _i using the homomorphic encryption algorithm Paillier's public key;

到

使用注册密钥

加密

将

发送给计算服务器CS，计算服务器CS加密索引I并发送给数据库DB；S103: Extractor extension registration vector

arrive

use registration key

encryption

Will

Send to the calculation server CS, the calculation server CS encrypts the index I and sends it to the database DB;

和E_K(T_A)发送给计算服务器CS；S104: The extractor expands the feature vector v _A to v′ _A , encrypts the extended feature vector v′ _A and the biometric template T _A , and

and E _K (T _A ) are sent to the computing server CS;

S105: The database DB transforms the received encrypted index I and the authentication query Q _A , calculates the similarity between each registration template and the authentication query Q _A , and sends the set of candidate templates to the calculation server CS, so that the calculation server CS can calculate Euclidean distance

S106: Support three update operations: add, delete and modify, ie new user registration, existing user revocation, and update of existing user keys and feature templates based on cancelable template modules.

The described light-weight biometric authentication method based on joint biometric identification includes a key generation phase, an encryption feature phase, an index generation phase, a token generation phase, an authentication phase and a feature update phase;

The key generation phase includes:

其中n＝pq，g是小于n²的随机数；私钥

其中α＝lcm(p-1，q-1)，

此外，可信中心TA为认证用户U_i基于对称加密算法AES生成对称加密密钥

(1) The trusted center TA generates two large prime numbers p and q for the authenticated user U _i , and generates a public key based on the homomorphic encryption algorithm Paillier

Where n=pq, g is a random number less than n ² ; the private key

where α = lcm(p-1, q-1),

In addition, the trusted center TA generates a symmetric encryption key for the authenticated user U _i based on the symmetric encryption algorithm AES

作为认证密钥，其中

最后，对于每个注册用户R_i，可信中心TA生成两个随机矩阵

作为索引构建密钥，其中

(2) First, the trusted center TA generates a random reversible matrix and its inverse matrix M for the authenticated user U _i , M ^-1 ∈ Z ^{2d ×2d} , where d is the dimension of the feature vector; then, for each authenticated user U _i , the trusted center TA generates two random matrices

As an authentication key, where

Finally, for each registered user R _i , the trusted center TA generates two random matrices

As an index build key, where

The encryption feature phase includes:

和一个n维向量v_f；定义操作

计算公式如下：(1) Obtain N M-dimensional vectors through face and fingerprint feature extraction

and an n-dimensional vector v _f ; define the operation

Calculated as follows:

The obfuscated biometric template obtained is as follows:

提取器基于“字符串连接”操作连接特征候选向量来构建d维生物特征注册向量，计算公式如下：(2) The extractor first randomly selects m ₁ (m ₁ ∈ M) numbers in each vector of v _B , obtains the relevant subscript data in each vector to construct the feature candidate vector v″ _i (i=1,… , N), the randomly generated subscript is defined as the user's registration key

The extractor constructs a d-dimensional biometric registration vector by concatenating feature candidate vectors based on the "string concatenation" operation, and the calculation formula is as follows:

v _R ＝v″ ₁ ||v″ ₂ ||…||v″ _N

获取每个向量中相关下标的数据来构造生物特征模板T_i，计算公式如下：(3) The extractor randomly selects m ₂ (m ₂ ∈ M) numbers in each vector of v _B , and the randomly generated subscript is defined as the template key of the user

Obtain the relevant subscript data in each vector to construct the biometric template T _i , the calculation formula is as follows:

(4) The extractor uses the homomorphic encryption algorithm Paillier's public key to encrypt the biometric template T _i , and the encryption formula is as follows:

The build index phase includes:

到

扩展公式如下：(1) First, the extractor expands each registration vector

arrive

The expansion formula is as follows:

是提取器为每个注册向量

随机选择的数字；in

is the extractor for each registration vector

randomly selected numbers;

加密

加密公式如下：(2) Then, the extractor uses the registration key

encryption

The encryption formula is as follows:

p₁>>p₂并且γ>>2|max(ε_i)|，

定义为从概率分布中随机选择的整数混淆向量；

是由注册向量

组成的密文；提取器以元组

的形式将

和

从注册用户R_i传输给计算服务器CS；当计算服务器CS收到所有注册用户的加密元祖时，将创建加密索引

其中U_max表示数据库DB中的用户总数；加密索引I将由计算服务器CS传输到数据库DB中进行存储。in,

p ₁ >>p ₂ and γ>>2|max(ε _i )|,

Defined as an integer confusion vector randomly selected from the probability distribution;

is given by the registration vector

The composed ciphertext; the extractor takes the tuple

in the form of

and

Transmitted from registered user R _i to computing server CS; when computing server CS receives encrypted tuples of all registered users, an encrypted index will be created

Where U _max represents the total number of users in the database DB; the encrypted index I will be transmitted by the computing server CS to the database DB for storage.

The token generation phase includes:

和模板密钥

的生物特征模板T_A；(1) First, the extractor extracts the feature vector v _A from the biometric feature of the authenticated user U _j and has the registration key

and the template key

biometric template T _A ;

(2) Then, the extractor expands the feature vector v _A to v′ _A , and the expansion formula is as follows:

是提取器为认证用户U_j的认证查询随机选择的数字，需要注意η_j是正数；in

is the number randomly selected by the extractor for the authentication query of the authenticated user U _j , it should be noted that η _j is a positive number;

对扩展向量v′_A进行加密，加密公式如下：(3) Next, the extractor uses the authentication key of the authenticated user U _j

Encrypt the extended vector v′ _A , the encryption formula is as follows:

是由提取器随机选择的整数混淆向量；提取器将加密的认证查询Q_A发送给计算服务器CS，然后再将认证查询Q_A发送给数据库DB进行认证；in

is an integer confusion vector randomly selected by the extractor; the extractor sends the encrypted authentication query Q _A to the computing server CS, and then sends the authentication query Q _A to the database DB for authentication;

对生物特征模版T_A进行加密以获得密文

除此之外，提取器使用认证用户U_j的对称密钥

对生物特征模版T_A进行加密以获得密文E_K(T_A)；提取器将

和E_K(T_A)发送给计算服务器CS，用于后续密文的欧几里德距离计算。(4) The extractor uses the public key of the homomorphic encryption algorithm Paillier of the authenticated user U _j

Encrypt the biometric template T _A to obtain the ciphertext

Among other things, the extractor authenticates the user U _j using the symmetric key

Encrypt the biometric template T _A to obtain the ciphertext E _K (T _A ); the extractor will

and E _K (T _A ) are sent to the computing server CS for Euclidean distance calculation of subsequent ciphertexts.

The certification phase includes:

汁算每个注册模板和认证查询Q_A之间的相似性；最后，数据库DB将候选模板集合发送给计算服务器CS，以便计算服务器CS使用与候选集和密文

相关的特征模板去计算两者之间的欧几里德距离；具体过程如下：The authentication process includes three steps: first, the database DB transforms the received encrypted index I and the authentication query Q _A ; then, the database DB transforms the received encrypted index I according to its stored encrypted index

Calculate the similarity between each registration template and the authentication query Q _A ; finally, the database DB sends the set of candidate templates to the computing server CS, so that the computing server CS uses the candidate set and the ciphertext

Related feature templates to calculate the Euclidean distance between the two; the specific process is as follows:

中的每一个

进行变换；然后，数据库DB再对从提取器接收到的认证查询Q_A进行变换，变换公式如下：(1) The database DB pairs the encrypted index received from the extractor

each of

Transformation; then, the database DB transforms the authentication query Q _A received from the extractor, and the transformation formula is as follows:

和在索引I中的每一个加密项

的相关性分数，计算公式如下：(2) Database DB calculates the transformed query

and every encrypted item in index I

The correlation score of is calculated as follows:

是相似性分数的随机数部分，消除

和

这两部分以获得上式的计算结果；in

is the random number part of the similarity score, eliminating

and

These two parts are used to obtain the calculation result of the above formula;

和k个候选模板之间的欧几里德距离；According to the calculation result, the database DB obtains the closest k index entries, and sends its corresponding biometric template set to the calculation server CS, and the calculation server CS will calculate the ciphertext template

and the Euclidean distance between k candidate templates;

对密文E_K(T_A)进行解密，以获取生物特征模版

然后，计算

和候选模板

之间的欧几里德距离，计算公式如下：(3) The computing server CS uses the symmetric key of the authenticated user U _i

Decrypt the ciphertext E _K (T _A ) to obtain the biometric template

Then, calculate

and candidate templates

The Euclidean distance between is calculated as follows:

If the Euclidean distance in the above formula is calculated according to its additive homomorphism under the homomorphic encryption algorithm Paillier, then the calculation result is as follows:

和

这两部分，利用加法同态将其转换为下面的式子，转换公式如下：for

and

These two parts are converted into the following formula by using additive homomorphism, and the conversion formula is as follows:

这部分，利用加法同态将其转换为下面的式子，转换公式如下：for

In this part, use additive homomorphism to convert it into the following formula, and the conversion formula is as follows:

进行转换，转换公式如下：Compute Server CS will

To convert, the conversion formula is as follows:

和加密的候选模板集合的前提下，计算服务器CS检查密文模板

和每个候选模板的欧几里德距离，结果中的最小欧几里德距离

是否满足设置的阈值T；如果小于设置的阈值T，计算服务器CS认为认证通过；否则，计算服务器CS认为认证失败。There is plaintext and ciphertext template T _A in computing server CS,

and the encrypted candidate template set, the computing server CS checks the ciphertext template

The Euclidean distance to each candidate template, the minimum Euclidean distance in the result

Whether it satisfies the set threshold T; if it is less than the set threshold T, the computing server CS considers the authentication passed; otherwise, the computing server CS considers the authentication failed.

The feature update phase includes:

The system supports three update operations: add, delete, and modify, namely new user registration, existing user revocation, and update of existing user keys and feature templates based on cancelable template modules. The specific process is as follows:

和加密的特征模板

提取器将

发送到计算服务器CS，计算服务器CS将其发送到数据库DB，数据库DB将

添加到存储的索引中以完成新用户的注册；(1) The new user U _ADD uploads his biometric data through the extractor, and the extractor processes the biometric data to generate encrypted registration vectors respectively

and the encrypted feature template

The extractor will

to the calculation server CS, the calculation server CS sends it to the database DB, and the database DB will

added to the stored index to complete the registration of a new user;

生成匹配索引条目

然后，提取器将索引条目

发送给计算服务器CS，计算服务器CS将其发送到数据库DB；最后，数据库DB在其存储的索引上删除索引条目

和匹配的加密模板

(2) The existing user revocation process requires three operations; first, the extractor collects and extracts the biometrics of the user U _DEL to be revoked; in addition, the extractor utilizes the registration key

generate matching index entries

The extractor will then index the entry

Sent to Compute Server CS, Compute Server CS sends it to Database DB; Finally, Database DB deletes the index entry on the index it stores

and the matching encryption template

和加密模板

此外，提取器生成新的注册和模板密钥

和

然后，提取器将

和

发送给计算服务器CS，计算服务器CS将其发送到数据库DB；最后，数据库DB利用索引项

去匹配加密索引I中与用户U_i相关的索引项

删除索引项

和相关的加密模板

并将

和

插入到加密索引I中，(3) When the biological template of user U _i is damaged or stolen, user U _i re-enters the biometric feature based on the cancelable template module; first, the extractor extracts the user biometric feature and obtains the registration index item

and encrypted template

Additionally, the extractor generates new registration and template keys

and

Then, the extractor will

and

Sent to the computing server CS, the computing server CS sends it to the database DB; finally, the database DB utilizes the index item

To match the index items related to user U _i in encrypted index I

delete index entry

and associated encryption templates

and will

and

Inserted into encrypted index I,

认证用户U_i的公钥；

认证用户U_i的私钥；

认证用户U_i的对称加密密钥；

认证密钥；R_i：注册用户；

索引构建密钥；v_B：生物特征模板；v_R：生物特征注册向量；T_i：候选模板；Paillier：一种同态加密算法；

注册向量；

扩展后的注册向量；I：加密索引；v_A：特征向量；v′_A：扩展后的特征向量；T_A：生物特征模版；Q_A：认证查询；

欧几里德距离；AES：一种对称加密算法；M：随机可逆矩阵；M^-1：M的逆矩阵；m₁、m₂：随机数；v″_i：特征候选向量；

注册用户R_i的注册密钥；

模板密钥；

生物特征模板T_i的加密形式；

注册向量

的加密形式；U_max：数据库DB中的用户总数；

生物特征模版T_A的同态加密形式；E_K(T_A)：生物特征模版T_A的对称加密形式；

认证用户U_j的对称密钥；

变换后的查询；U_ADD：新用户；

新用户U_ADD加密的注册向量；

新用户U_ADD加密的特征模板；U_DEL：待撤销用户；

待撤销用户U_DEL的注册密钥；

与待撤销用户U_DEL匹配的索引条目；

与待撤销用户U_DEL匹配的加密模板；

用户U_i新的注册索引项；

用户U_i新的加密模板；

用户U_i新的注册密钥；

用户U_i新的模板密钥；

新定义的向量运算；||：字符串连接操作；∑：累加运算；Π：连乘运算。Among them, TA: trusted center; CS: computing server; DB: database; U _i , U _j : authenticated users;

Authenticating the public key of user U _i ;

Authenticating the private key of user U _i ;

symmetric encryption key for authenticating user U _i ;

Authentication key; R _i : registered user;

Index construction key; v _B : biometric template; v _R : biometric registration vector; T _i : candidate template; Paillier: a homomorphic encryption algorithm;

registration vector;

Extended registration vector; I: encrypted index; v _A : feature vector; v′ _A : extended feature vector; T _A : biometric template; Q _A : authentication query;

Euclidean distance; AES: a symmetric encryption algorithm; M: random reversible matrix; M ^-1 : the inverse matrix of M; m ₁ , m ₂ : random numbers; v″ _i : feature candidate vector;

Registration key of registered user R _i ;

template key;

An encrypted form of the biometric template T _i ;

registration vector

encrypted form; U _max : the total number of users in the database DB;

The homomorphic encryption form of the biometric template T _A ; E _K (T _A ): the symmetric encryption form of the biometric template T _A ;

symmetric key for authenticating user U _j ;

Transformed query; U _ADD : new user;

New user U _ADD encrypted registration vector;

Feature template encrypted by new user U _ADD ; U _DEL : user to be revoked;

The registration key of the user U _DEL to be revoked;

Index entries matching the user U _DEL to be revoked;

The encryption template matching the user U _DEL to be revoked;

User U _i 's new registration index item;

User U _i 's new encryption template;

User U _i new registration key;

User U _i new template key;

Newly defined vector operations; ||: string concatenation operation; ∑: accumulation operation; Π: multiplication operation.

The lightweight biometric authentication method based on combined biometrics is stored in a program storage medium that receives user input, and is executed by electronic equipment through a computer program.

The lightweight biometric authentication method based on joint biometrics is implemented using a lightweight biometric authentication system, which includes:

An extractor for extracting biometrics, generating cancelable template modules, and encrypting templates;

a trusted center for generating keys;

Calculation server for Euclidean distance calculation of encrypted biometric features;

Database for storing indexes.

The lightweight biometric authentication system is carried on a terminal, and the terminal is an Internet of Things terminal.

The beneficial effects of the present invention are: the present invention uses the newly proposed random bit to generate RBG and encryption process to construct biometric templates and related indexes, which can protect the privacy of outsourced stored biometric templates and the confidentiality of the identity authentication process; After extracting the authentication template, use random bits to generate RBG and matrix keys to confuse and encrypt the template to obtain a token, which can ensure the security of the entire authentication process and the incompatibility of the query; the authentication matching process screens out the tokens that are close to the token Matching set, and then compare the similarity between the authentication template and the template in the matching set one by one, and perform similarity calculation based on the newly proposed encrypted vector distance calculation method, so that the authentication process is very robust and can ensure the accuracy of authentication; Therefore, the method of the present invention achieves higher safety and accuracy at a lower cost.

The present invention adopts low-overhead-based encryption process and random bit generation to construct biometric templates and related indexes, and users will generate and obtain keys from random bits as identity authentication passwords, which is the basis for subsequent realization of joint knowledge and biometric identity authentication; After the extractor extracts the authentication template, it uses random bit generation and matrix key to confuse and encrypt the template to obtain tokens, and uses searchable encryption technology to design a retrieval method based on tokens and encrypted indexes to retrieve the k most The template index close to the authentication template can ensure the security of the entire authentication process and the non-connectability of the query; the present invention proposes a biometric template construction method combining face and fingerprint, which uses local binary features LBP and based on The fingerprint characteristics of the details, the matching set close to the token is screened out during the authentication process, and then the similarity between the authentication template and the template in the matching set is compared one by one, and the similarity calculation is performed based on the newly proposed encryption vector distance calculation method, so that the authentication process has Strong robustness can ensure the accuracy of authentication. The invention satisfies the confidentiality, renewability, revocability, irreversibility and non-connectability of templates in biological feature identification, and realizes the balance between low cost of biological features and high identification security requirements.

The present invention is compared and analyzed with the classical biometric authentication method. The security comparison results are shown in Table 1, where "√" indicates that the security requirement is met, "×" indicates that the security requirement is not met, and "*" indicates that the security requirement is partially met.

Table 1 Security comparison

In Table 1, the method of the present invention has renewability and revocability that other classical biometric authentication methods do not have; the method of Zhu et al. has verifiability and collusion resistance, but due to the introduction of bilinear pairing, which inevitably produces high system overhead, making this method unsuitable for mobile devices. The method of the invention realizes the balance between low cost of biometric authentication and high security requirements.

Description of drawings

Fig. 1 is a flowchart of a lightweight biometric authentication method according to an embodiment of the present invention.

Fig. 2 is a system frame diagram of a lightweight biometric authentication system according to an embodiment of the present invention.

Fig. 3 is a flow chart of implementing a lightweight biometric authentication method according to an embodiment of the present invention.

Fig. 4, Fig. 5 and Fig. 6 are comparison charts of authentication accuracy and authentication time in different databases (ORL, Yale and FERET databases) between the lightweight biometric authentication method of the embodiment of the present invention and other classic biometric authentication methods.

Fig. 7, Fig. 8 and Fig. 9 are the simulation diagrams of the authentication accuracy of the LBP feature descriptor dimension in different databases (ORL, Yale and FERET databases) for different data volumes in the lightweight biometric authentication method of the embodiment of the present invention.

Fig. 10 is a comparison diagram of the time cost of key generation and token generation in the light-weight biometric authentication method of the embodiment of the present invention and other classic biometric authentication methods under different feature vector sizes.

Fig. 11 is a comparison diagram of the time cost of index construction in the FERET database between the lightweight biometric authentication method and other classic biometric authentication methods according to the embodiment of the present invention.

Fig. 12 is a comparison diagram of the time cost of querying in the FERET database between the lightweight biometric authentication method and other classic biometric authentication methods according to the embodiment of the present invention.

Fig. 13, Fig. 14 and Fig. 15 are comparison diagrams of the time cost of encryption vector calculation between the lightweight biometric authentication method of the embodiment of the present invention and other classic biometric authentication methods under different vector sizes and data volumes.

Detailed ways

The present invention will be described in further detail below in conjunction with the embodiments and accompanying drawings. It should be understood that the specific embodiments described here are only used to explain the present invention, not to limit the present invention.

As shown in Figure 1, a lightweight biometric authentication method based on joint biometrics includes the following steps:

私钥

对称加密密钥

和认证密钥

为注册用户R_i生成索引构建密钥

S101: The trusted center TA generates a series of keys, and generates a public key for the authenticated user U _i

private key

Symmetric encryption key

and authentication key

Generate an index build key for registered user R _i

S102: The extractor constructs a d-dimensional biometric registration vector v _R and a biometric template T _i by confusing the biometric template v _B , and encrypts the biometric template T _i using the homomorphic encryption algorithm Paillier's public key;

到

使用注册密钥

加密

将

发送给计算服务器CS，计算服务器CS加密索引I并发送给数据库DB；S103: Extractor extension registration vector

arrive

use registration key

encryption

Will

Send to the calculation server CS, the calculation server CS encrypts the index I and sends it to the database DB;

和E_K(T_A)发送给计算服务器CS；S104: The extractor expands the feature vector v _A to v′ _A , encrypts the extended feature vector v′ _A and the biometric template T _A , and

and E _K (T _A ) are sent to the computing server CS;

S105: The database DB transforms the received encrypted index I and the authentication query Q _A , calculates the similarity between each registration template and the authentication query Q _A , and sends the set of candidate templates to the calculation server CS, so that the calculation server CS can calculate Euclidean distance

S106: Support three update operations: add, delete and modify, ie new user registration, existing user revocation, and update of existing user keys and feature templates based on cancelable template modules.

As shown in Figure 2, the lightweight biometric authentication method based on joint biometrics is implemented using a lightweight biometric authentication system, and the lightweight biometric authentication system includes:

Extractor: As a fully trusted entity in the system, the extractor has sufficient computing power but does not have a large storage space. It is mainly responsible for extracting biometric features, generating multi-mode cancelable templates together with the trusted center, and according to the trusted Key encryption template assigned by the center;

Trusted center: responsible for assisting the extractor in generating multi-mode cancelable templates, and generating different template encryption keys for different users;

Computing server: There are multiple trusted computing servers in the system. Each computing server uses powerful computing power to provide services for all users in the system sub-area. The computing server is responsible for the Euclidean distance calculation of encrypted biometrics, and according to The calculation result between the candidate template set and the authentication template returns the final authentication result;

Database: As the entity with the strongest computing power and storage space in the system, the distributed database can store the biometric templates of multiple users, and associate user identities and feature templates by establishing encrypted query indexes; the database is a semi-trusted Entities that will fully execute instructions and perform statistical analysis on stored information.

As shown in Figure 3, the described lightweight biometric authentication method based on joint biometrics includes a key generation phase, an encryption feature phase, an index generation phase, a token generation phase, an authentication phase and a feature update phase;

The key generation phase includes:

其中n＝pq，g是小于n²的随机数；私钥

其中α＝lcm(p-1，q-1)，

此外，可信中心TA为认证用户U_i基于对称加密算法AES生成对称加密密钥

(1) The trusted center TA generates two large prime numbers p and q for the user U _i , and generates a public key based on the homomorphic encryption algorithm Paillier

Where n=pq, g is a random number less than n ² ; the private key

where α = lcm(p-1, q-1),

In addition, the trusted center TA generates a symmetric encryption key for the authenticated user U _i based on the symmetric encryption algorithm AES

作为认证密钥，其中

最后，对于每个注册用户R_i，可信中心TA生成两个随机矩阵

作为索引构建密钥，其中

(2) First, the trusted center TA generates a random reversible matrix and its inverse matrix M for the authenticated user U _i , M ^-1 ∈ Z ^{2d ×2d} , where d is the dimension of the feature vector; then, for each authenticated user U _i , the trusted center TA generates two random matrices

As an authentication key, where

Finally, for each registered user R _i , the trusted center TA generates two random matrices

As an index build key, where

The encryption feature phase includes:

和一个n维向量v_f；定义操作

计算公式如下：(1) Through face and fingerprint feature extraction, N M-dimensional vectors can be obtained

and an n-dimensional vector v _f ; define the operation

Calculated as follows:

The obfuscated biometric template obtained is as follows:

提取器基于“字符串连接”操作连接特征候选向量来构建d维生物特征注册向量，计算公式如下：(2) The extractor first randomly selects m ₁ (m ₁ ∈ M) numbers in each vector of v _B , obtains the relevant subscript data in each vector to construct the feature candidate vector v″ _i (i=1,… , N), the randomly generated subscript is defined as the user's registration key

The extractor constructs a d-dimensional biometric registration vector by concatenating feature candidate vectors based on the "string concatenation" operation, and the calculation formula is as follows:

v _R ＝v″ ₁ ||v″ ₂ ||…||v″ _N

获取每个向量中相关下标的数据来构造生物特征模板T_i，计算公式如下：(3) The extractor randomly selects m ₂ (m ₂ ∈ M) numbers in each vector of v _B , and the randomly generated subscript is defined as the template key of the user

Obtain the relevant subscript data in each vector to construct the biometric template T _i , the calculation formula is as follows:

(4) The extractor uses the homomorphic encryption algorithm Paillier's public key to encrypt the biometric template T _i , and the encryption formula is as follows:

The build index phase includes:

到

扩展公式如下：(1) First, the extractor expands each registration vector

arrive

The expansion formula is as follows:

是提取器为每个注册向量

随机选择的数字；in

is the extractor for each registration vector

randomly selected numbers;

加密

加密公式如下：(2) Then, the extractor uses the registration key

encryption

The encryption formula is as follows:

p₁>>p₂并且γ>>2|max(ε_i)|，

定义为从概率分布中随机选择的整数混淆向量；在随后的认证阶段中，上述参数将用于向量相似性计算；

是由注册向量

组成的密文；提取器以元组

的形式将

和

从注册用户R_i传输给计算服务器CS；当计算服务器CS收到所有注册用户的加密元祖时，将创建加密索引

其中U_max表示数据库DB中的用户总数；加密索引I将由计算服务器CS传输到数据库DB中进行存储。in,

p ₁ >>p ₂ and γ>>2|max(ε _i )|,

Defined as an integer confusion vector randomly selected from a probability distribution; in the subsequent authentication phase, the above parameters will be used for vector similarity calculations;

is given by the registration vector

The composed ciphertext; the extractor takes the tuple

in the form of

and

Transmitted from registered user R _i to computing server CS; when computing server CS receives encrypted tuples of all registered users, an encrypted index will be created

Where U _max represents the total number of users in the database DB; the encrypted index I will be transmitted by the computing server CS to the database DB for storage.

The token generation phase includes:

和模板密钥

的生物特征模板T_A；(1) First, the extractor extracts the feature vector v _A from the biometric feature of the authenticated user U _j and has the registration key

and the template key

biometric template T _A ;

(2) Then, the extractor expands the feature vector v _A to v′ _A , and the expansion formula is as follows:

是提取器为认证用户U_j的认证查询随机选择的数字，需要注意η_j是正数；in

is the number randomly selected by the extractor for the authentication query of the authenticated user U _j , it should be noted that η _j is a positive number;

对扩展向量v′_A进行加密，加密公式如下：(3) Next, the extractor uses the authentication key of the authenticated user U _j

Encrypt the extended vector v′ _A , the encryption formula is as follows:

是由提取器随机选择的整数混淆向量；提取器将加密的认证查询Q_A发送给计算服务器CS，然后再将认证查询Q_A发送给数据库DB进行认证；in

is an integer confusion vector randomly selected by the extractor; the extractor sends the encrypted authentication query Q _A to the computing server CS, and then sends the authentication query Q _A to the database DB for authentication;

对生物特征模版T_A进行加密以获得密文

除此之外，提取器使用认证用户U_j的对称密钥

对生物特征模版T_A进行加密以获得密文E_K(T_A)；提取器将

和E_K(T_A)发送给计算服务器CS，用于后续密文的欧几里德距离计算。(4) The extractor uses the public key of the homomorphic encryption algorithm Paillier of the authenticated user U _j

Encrypt the biometric template T _A to obtain the ciphertext

Among other things, the extractor authenticates the user U _j using the symmetric key

Encrypt the biometric template T _A to obtain the ciphertext E _K (T _A ); the extractor will

and E _K (T _A ) are sent to the computing server CS for Euclidean distance calculation of subsequent ciphertexts.

The certification phase includes:

计算每个注册模板和认证查询Q_A之间的相似性；最后，数据库DB将候选模板集合发送给计算服务器CS，以便计算服务器CS使用与候选集和密文

相关的特征模板去计算两者之间的欧几里德距离；具体过程如下：The authentication process includes three steps: first, the database DB transforms the received index I and the authentication query Q _A ; then, the database DB transforms the received index I according to the encrypted index

Calculate the similarity between each registration template and the authentication query Q _A ; finally, the database DB sends the set of candidate templates to the calculation server CS, so that the calculation server CS uses the candidate set and the ciphertext

Related feature templates to calculate the Euclidean distance between the two; the specific process is as follows:

中的每一个

进行变换；然后，数据库DB再对从提取器接收到的认证查询Q_A进行变换，变换公式如下：(1) The database DB pairs the encrypted index received from the extractor

each of

Transformation; then, the database DB transforms the authentication query Q _A received from the extractor, and the transformation formula is as follows:

和在加密索引I中的每一个加密项

的相关性分数，计算公式如下：(2) Database DB calculates the transformed query

and each encrypted item in the encrypted index I

The correlation score of is calculated as follows:

是相似性分数的随机数部分，需要注意的是由于p₁>>p₂且

所以

和

这两部分的值无限趋近于0；由于计算结果在域

上四舍五入，因此可以消除

和

这两部分以获得上式的计算结果；in

is the random number part of the similarity score, it should be noted that since p ₁ >>p ₂ and

so

and

The values of these two parts are infinitely close to 0; since the calculation results are in the domain

is rounded up, so it can be eliminated

and

These two parts are used to obtain the calculation result of the above formula;

和k个候选模板之间的欧几里德距离；According to the calculation result, the database DB obtains the closest k index entries, and sends its corresponding biometric template set to the calculation server CS, and the calculation server CS will calculate the ciphertext template

and the Euclidean distance between k candidate templates;

对密文E_K(T_A)进行解密，以获取生物特征模版

然后，计算

和候选模板

之间的欧几里德距离，计算公式如下：(3) The computing server CS uses the symmetric key of the authenticated user U _i

Decrypt the ciphertext E _K (T _A ) to obtain the biometric template

Then, calculate

and candidate templates

The Euclidean distance between is calculated as follows:

If the Euclidean distance in the above formula is calculated according to its additive homomorphism under the homomorphic encryption algorithm Paillier, then the calculation result is as follows:

和

这两部分，可以利用加法同态将其转换为下面的式子，转换公式如下：for

and

These two parts can be converted into the following formula by using additive homomorphism. The conversion formula is as follows:

这部分，可以利用加法同态将其转换为下面的式子，转换公式如下：for

This part can be converted into the following formula by using additive homomorphism. The conversion formula is as follows:

进行转换，转换公式如下：Due to the conversion formula above, the calculation server CS will

To convert, the conversion formula is as follows:

和加密的候选模板集合的前提下，计算服务器CS可以检查密文模板

和每个候选模板的欧几里德距离，结果中的最小欧几里德距离

是否满足设置的阈值T；如果小于设置的阈值T，计算服务器CS认为认证通过；否则，计算服务器CS认为认证失败。There is plaintext and ciphertext template T _A in computing server CS,

and the encrypted candidate template set, the computing server CS can check the ciphertext template

The Euclidean distance to each candidate template, the minimum Euclidean distance in the result

Whether it satisfies the set threshold T; if it is less than the set threshold T, the computing server CS considers the authentication passed; otherwise, the computing server CS considers the authentication failed.

The feature update phase includes:

The system supports three update operations: add, delete, and modify, namely new user registration, existing user revocation, and update of existing user key and feature templates based on the cancelable template module. The specific process is as follows:

和加密的特征模板

提取器将

发送到计算服务器CS，计算服务器CS将其发送到数据库DB，数据库DB将

添加到存储的加密索引中以完成新用户的注册；(1) The new user U _ADD uploads his biometric data through the extractor, and the extractor processes the biometric data to generate encrypted registration vectors respectively

and the encrypted feature template

The extractor will

to the calculation server CS, the calculation server CS sends it to the database DB, and the database DB will

Added to the stored encrypted index to complete the registration of new users;

生成匹配索引条目

然后，提取器将索引条目

发送给计算服务器CS，计算服务器CS将其发送到数据库DB；最后，数据库DB在其存储的索引上删除索引条目

和匹配的加密模板

(2) The existing user revocation process requires three operations; first, the extractor collects and extracts the biometrics of the user U _DEL to be revoked; in addition, the extractor utilizes the registration key

generate matching index entries

The extractor will then index the entry

Sent to Compute Server CS, Compute Server CS sends it to Database DB; Finally, Database DB deletes the index entry on the index it stores

and the matching encryption template

和加密模板

此外，提取器生成新的注册和模板密钥

和

然后，提取器将

和

发送给计算服务器CS，计算服务器CS将其发送到数据库DB；最后，数据库DB利用索引项

去匹配索引I中与用户U_i相关的索引项

删除索引项

和相关的加密模板

并将

和

插入到加密索引I中。(3) When the biological template of user U _i is damaged or stolen, user U _i can re-enter the biometric feature based on the cancelable template module; first, the extractor extracts the user biometric feature and obtains the registration index item

and encrypted template

Additionally, the extractor generates new registration and template keys

and

Then, the extractor will

and

Sent to the computing server CS, the computing server CS sends it to the database DB; finally, the database DB utilizes the index item

To match the index items related to user U _i in index I

delete index entry

and associated encryption templates

and will

and

Insert into encrypted index I.

认证用户U_i的公钥；

认证用户U_i的私钥；

认证用户U_i的对称加密密钥；

认证密钥；R_i：注册用户；

索引构建密钥；v_B：生物特征模板；v_R：生物特征注册向量；T_i：候选模板；Paillier：一种同态加密算法；

注册向量；

扩展后的注册向量；I：加密索引；v_A：特征向量；v′_A：扩展后的特征向量；T_A：生物特征模版；Q_A：认证查询；

欧几里德距离；AES：一种对称加密算法；M：随机可逆矩阵；M^-1：M的逆矩阵；m₁、m₂：随机数；v″_i：特征候选向量；

注册用户R_i的注册密钥；

模板密钥；

生物特征模板T_i的加密形式；

注册向量

的加密形式；U_max：数据库DB中的用户总数；

生物特征模版T_A的同态加密形式；E_K(T_A)：生物特征模版T_A的对称加密形式；

认证用户U_j的对称密钥；

变换后的查询；U_ADD：新用户；

新用户U_ADD加密的注册向量；

新用户U_ADD加密的特征模板；U_DEL：待撤销用户；

待撤销用户U_DEL的注册密钥；

与待撤销用户U_DEL匹配的索引条目；

与待撤销用户U_DEL匹配的加密模板；

用户U_i新的注册索引项；

用户U_i新的加密模板；

用户U_i新的注册密钥；

用户U_i新的模板密钥；

新定义的向量运算；||：字符串连接操作；∑：累加运算；Π：连乘运算。Among them, TA: trusted center; CS: computing server; DB: database; U _i , U _j : authenticated users;

Authenticating the public key of user U _i ;

Authenticating the private key of user U _i ;

symmetric encryption key for authenticating user U _i ;

Authentication key; R _i : registered user;

Index construction key; v _B : biometric template; v _R : biometric registration vector; T _i : candidate template; Paillier: a homomorphic encryption algorithm;

registration vector;

Extended registration vector; I: encrypted index; v _A : feature vector; v′ _A : extended feature vector; T _A : biometric template; Q _A : authentication query;

Euclidean distance; AES: a symmetric encryption algorithm; M: random reversible matrix; M ^-1 : the inverse matrix of M; m ₁ , m ₂ : random numbers; v″ _i : feature candidate vector;

Registration key of registered user R _i ;

template key;

An encrypted form of the biometric template T _i ;

registration vector

encrypted form; U _max : the total number of users in the database DB;

The homomorphic encryption form of the biometric template T _A ; E _K (T _A ): the symmetric encryption form of the biometric template T _A ;

symmetric key for authenticating user U _j ;

Transformed query; U _ADD : new user;

New user U _ADD encrypted registration vector;

Feature template encrypted by new user U _ADD ; U _DEL : user to be revoked;

The registration key of the user U _DEL to be revoked;

Index entries matching the user U _DEL to be revoked;

The encryption template matching the user U _DEL to be revoked;

User U _i 's new registration index item;

User U _i 's new encryption template;

User U _i new registration key;

User U _i new template key;

Newly defined vector operations; ||: string concatenation operation; ∑: accumulation operation; Π: multiplication operation.

In order to verify the usability of the present invention, the test results of the lightweight biometric authentication method SELBA based on combined biometrics under simulation will be shown and explained below, and the simulation environment: on a PC with a CPU 2.10GHz, Windows environment.

Figure 4, Figure 5 and Figure 6 are the comparisons of authentication accuracy and authentication time between the lightweight biometric authentication method SELBA based on joint biometrics and other classic biometric authentication methods in different databases (ORL, Yale, and FERET databases). The results show that the accuracy of the method of the present invention in different databases is lower than that based on CNN, and higher than that based on Gabor and PCA. When the amount of data in different databases reaches a certain level, the authentication accuracy of the method of the invention can be stabilized above 95%. Under different methods, the time consumption will increase linearly with the increase of data volume. The time consumption of the method of the present invention is slightly higher than that of the Gabor-based and PCA-based methods, but the time consumption of the CNN-based method is about four times that of other methods. High-accuracy methods are bound to be accompanied by high overhead. Although the accuracy based on CNN is the highest, the application on mobile devices needs to consider the accuracy and efficiency of identity authentication. Compared with other methods, the method of the invention realizes the balance between authentication low cost and high safety requirement.

Figure 7, Figure 8 and Figure 9 are the authentication accuracy of different data volumes in different databases (ORL, Yale and FERET databases) for the dimensions of the LBP feature descriptor in the lightweight biometric authentication method SELBA based on joint biometrics Effects are simulated. The results show that in different databases, the larger the data volume, the higher the recognition accuracy. In addition, the larger the dimensionality of the descriptor, the higher the recognition accuracy. It can be seen from Fig. 7, Fig. 8 and Fig. 9 that the recognition rate of the 3×3 dimensional descriptor is more than 10% higher than that of the 8×6 dimensional descriptor. According to the simulation results, it can be observed that, compared with the other three methods, the method of the present invention maintains a high identity authentication accuracy on the basis of protecting the privacy of biological data, and does not cause large additional calculation overhead.

Figure 10 is a comparison of the time of key generation and token generation under different eigenvector sizes between the lightweight biometric authentication method SELBA based on joint biometric identification and other classic biometric authentication methods. The results show that the time cost of key generation for the method of the present invention and the method of Zhu et al. increases linearly with the increase of vector size. When the vector size is 16 bits, the average time to create the key is about 110 ms and 290 ms, respectively. When the size increases to 256 bits, their average time overhead increases to 1100ms and 1253ms, respectively. However, for the method of Zhou et al., when the vector size increases from 16 bits to 256 bits, the average time cost increases from 100ms to 2300ms, showing an exponential growth trend. The method of Zhou et al. uses a matrix as a key, and the size of the matrix will change synchronously with the size of the vector, while the method of the present invention and the method of Zhu et al. are based on homomorphic encryption, so the size of the key can not be affected by the size of the vector. In addition, the time cost of token generation in the three methods also increases with the increase of vector dimension, but the increase is much smaller than that of key generation. When the size is 16 bits, the average token generation time of the method of the present invention, the method of Zhu et al. and the method of Zhou et al. is about 30ms, 220ms and 60ms respectively. When the size was increased to 256 bits, the average times reached around 150ms, 1200ms and 860ms respectively. Due to the bilinear pairing introduced by the method of Zhu et al., its overhead is significantly higher than the other two methods during token generation. Simulation results show that the method of the present invention has a low time cost for key generation and token generation, and has advantages in practice compared with other biometric authentication methods.

Figure 11 is a time comparison of index construction in the FERET database between the lightweight biometric authentication method SELBA based on joint biometrics and other classic biometric authentication methods. The index structures of these three methods are all inverted indexes, and the time cost of index construction increases linearly with the increase of the amount of training data. When the amount of training data is 200, the average time overhead of the method of the present invention, the method of Zhu et al. and the method of Zhou et al. is kept at about 0.16s, 0.45s and 0.38s respectively. When the amount of data increases to 1000, the cost increases to about 10 seconds, 12 seconds and 9 seconds. Simulation results show that in practical applications, the time overhead of the index construction of the method of the present invention is within an acceptable range.

Figure 12 is a comparison of query time in the FERET database between the lightweight biometric authentication method SELBA based on joint biometrics and other classic biometric authentication methods. The query time of the method of the present invention, the method of Zhu et al. and the method of Zhou et al. increases exponentially with the increase of the vector size. When the vector size is 16 bits, the variation of the training data has little impact on the query time cost, while when the vector size is 64 or 256 bits, the variation of the training data will greatly affect the query overhead. However, smaller eigenvector sizes do not describe features well. In practical application scenarios, setting the feature vector to 64 bits is a better choice considering accuracy and efficiency. Simulation results show that, compared with other methods, the method of the present invention achieves a balance between authentication low cost and high security requirements.

Fig. 13, Fig. 14 and Fig. 15 are comparisons of the time overhead of encryption vector calculation under different vector sizes and data volumes between the lightweight biometric authentication method SELBA based on joint biometrics and other classic biometric authentication methods. Figure 13 and Figure 14 show the variation of time overhead as the amount of data increases when the vector size is 16 bits and 64 bits, respectively. Among them, the cost of the method of Zhu et al. is much higher than the method of the present invention and the method of Zhou et al. This is mainly due to the fact that the calculation process of Zhu et al.'s method contains many linear pairing operations, and the validity of tokens needs to be verified before calculation. Figure 15 shows that when the vector size increases to 256 bits, the time cost of the method of the present invention and the method of Zhou et al. under different data volumes gradually approach the method of Zhu et al. Simulation results show that, compared with other methods, the method of the present invention achieves a balance between authentication low cost and high security requirements.

It should be noted that the method of the present invention can be realized by hardware, software or a combination of software and hardware, and the hardware part can be realized by using dedicated logic; the software part can be stored in a memory, and executed by an appropriate instruction system, such as a microprocessor or a dedicated Design the hardware to execute. Those of ordinary skill in the art will understand that the above-described devices and methods can be implemented using computer-executable instructions and/or contained in processor control code, for example, on a carrier medium such as a magnetic disk, CD or DVD-ROM, such as a read-only memory Such code is provided on a programmable memory (firmware) or on a data carrier such as an optical or electronic signal carrier. The device and its modules of the present invention may be implemented by hardware circuits such as VLSI or gate arrays, semiconductors such as logic chips, transistors, etc., or programmable hardware devices such as field programmable gate arrays, programmable logic devices, etc., It can also be realized by software executed by various types of processors, or by a combination of the above-mentioned hardware circuits and software such as firmware.

The above is only an embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Anyone familiar with the technical field within the technical scope disclosed in the present invention, any work done within the spirit and principles of the present invention Modifications, equivalent replacements and improvements, etc., should all be covered within the protection scope of the present invention.

Claims (10)

1. A lightweight biometric authentication method based on joint biometric identification is characterized by comprising the following steps:

s101: the trusted center TA generates a series of keys for authenticating the user U _i Generating public keys

Private key

Symmetric encryption key

And an authentication key

For registering a user R _i Generating an index build key

S102: extractor by obfuscating biometric templates v _B To construct a d-vitamin profile registryQuantity v _R And a biometric template T _i And encrypting the biological characteristic template T by using a public key of a homomorphic encryption algorithm Paillier _i ；

S103: extractor extended registration vector

To

Using registration keys

Encryption

Will be provided with

Sending the index I to a computing server CS, encrypting the index I by the computing server CS and sending the index I to a database DB;

s104: extractor extended feature vector v _A To v' _A For extended feature vector v' _A And a biometric template T _A Is encrypted and will

And E _K (T _A ) Sending to a computing server CS;

s105: the database DB receives the encryption index I and the authentication query Q _A Transforming, computing each enrollment template and authentication query Q _A The similarity between the candidate templates is sent to the computing server CS so that the computing server CS can compute the Euclidean distance

S106: three update operations are supported: addition, deletion, and modification, i.e., new user registration, existing user revocation, and updating of existing user keys and feature templates based on the revocable template module.

2. The method of claim 1, comprising a key generation phase, a feature encryption phase, an index generation phase, a token generation phase, an authentication phase, and a feature update phase;

the key generation phase comprises:

(1) The trusted center TA is an authenticated user U _i Generating two large prime numbers p and q, and generating a public key based on a homomorphic encryption algorithm Paillier

Wherein n = pq, g is a random number less than n 2; private key

Wherein α = lcm (p-1, q-1),

in addition, the trusted center TA is an authenticated user U _i Symmetric encryption key generation based on symmetric encryption algorithm AES

(2) Firstly, the trusted center TA is the authenticated user U _i Generating a random invertible matrix and its inverses M, M ^-1 ∈Z ^2d×2d Where d is the dimension of the feature vector; then, for each authenticated user U _i The trusted center TA generates two random matrices

As an authentication key, wherein

Finally, for each registered user R _i The trusted center TA generates two random matrices

Constructing the key as an index, wherein

3. The method of claim 2, wherein the encryption characterization phase comprises:

(1) Obtaining N M-dimensional vectors through face and fingerprint feature extraction

And an n-dimensional vector v _f I =1, \ 8230;, N; defining operations

The calculation formula is as follows:

the obfuscated biometric templates obtained were as follows:

(2) The extractor is first operated at v _B Randomly selects m in each vector ₁ Number, m ₁ E is M, and data of relevant subscripts in each vector are obtained to construct a characteristic candidate vector v _i ", i =1, \ 8230;, N, randomly generated subscript defines the user's registration key

The extractor connects the feature candidate vectors based on the 'string connection' operation to construct a d-vitamin feature registration vector, and the calculation formula is as follows:

v _R ＝v ₁ ″||v ₂ ″||…||v _N ″

l |: character string connection operation;

(3) Extractor at v _B Randomly selects m in each vector ₂ Number m ₂ Belongs to M, randomly generated subscript is defined as a template key of a user

Obtaining data of related subscript in each vector to construct biological feature template T _i The calculation formula is as follows:

(4) The extractor encrypts the biological characteristic template Ti by using a public key of a homomorphic encryption algorithm Paillier, wherein an encryption formula is as follows:

4. the method of claim 3, wherein generating the index phase comprises:

(1) First, the extractor expands each registration vector

To

The expansion formula is as follows:

wherein

Is a liftExtractor for each registration vector

A randomly selected number; sigma: performing accumulation operation;

(2) The extractor then uses the registration key

Encryption

The encryption formula is as follows:

wherein,

p ₁ >>p ₂ and gamma is>>2|max(ε _i )|，

Defined as an integer confusion vector randomly selected from a probability distribution;

is formed by a registration vector

A composed ciphertext; extractor with tuples

In the form of

And

from registered user R _i Transmitting to a computing server CS; when the computing server CS receives the encrypted metanodes of all registered users, an encryption index will be created

Wherein U is _max Representing the total number of users in the database DB; the encryption index I will be transmitted by the calculation server CS to the database DB for storage.

5. The method of claim 4, wherein the generating the token phase comprises:

(1) First, the extractor authenticates the user U from the certificate _i Extracting a feature vector v from the biological features _A And has a registration key

And template key

Biological characteristic template T _A ；

(2) The extractor then expands the feature vector v _A To v' _A The expansion formula is as follows:

wherein

The extractor is an authenticated user U _j To authenticate the randomly selected number, needs to pay attention to eta _j Is a positive number;

(3) Next, the extractor uses the authenticated user U _j Authentication key of

To extension vector v' _A Encryption is carried out, and an encryption formula is as follows:

wherein

Is an integer confusion vector randomly selected by the extractor; the extractor will encrypt the authentication query Q _A Sends it to the computing server CS, and then sends an authentication query Q _A Sending the data to a database DB for authentication;

(4) Extractor use authentication user U _j The public key of the homomorphic encryption algorithm Paillier

Template T for biological characteristics _A Encrypting to obtain ciphertext

In addition, the extractor uses the authenticated user U _j Symmetric key of (2)

To biological characteristic template T _A Encrypt to obtain ciphertext E _K (T _A ) (ii) a The extractor will

And E _K (T _A ) And sending the result to a computing server CS for Euclidean distance computation of subsequent ciphertexts.

6. The method of claim 5, wherein the authentication process comprises three steps: first, the database DB receives the encryption index I and the authentication query Q _A Carrying out conversion; the database DB then stores the encryption index according to it

Computing each enrollment template and authentication query Q _A Similarity between them; finally, the database DB sends the set of candidate templates to the computation server CS for use by the computation server CS with the set of candidate templates and the ciphertext

The correlated characteristic template calculates the Euclidean distance between the two characteristic templates; the specific process is as follows:

(1) Database DB for encryption index received from extractor

Each of which is

Carrying out conversion; the database DB then re-matches the authentication query Q received from the extractor _A And performing transformation, wherein the transformation formula is as follows:

(2) Database DB computation-transformed queries

And each encrypted item in the encryption index I

The formula for calculating the relevance score is as follows:

wherein

Is the random number part of the similarity score, eliminated

And

these two parts are used to obtain the calculation result of the above formula;

according to the calculation result, the database DB obtains the nearest k index entries, sends the corresponding biological characteristic template set to the calculation server CS, and the calculation server CS calculates the ciphertext template

And euclidean distances between the k candidate templates;

(3) Computing server CS uses authentication user U _i Symmetric key of

For ciphertext E _K (T _A ) Decrypting to obtain the biological characteristic template

Then, calculate

And candidate templates

The euclidean distance between them, the calculation formula is as follows:

if the Euclidean distance in the above equation is computed from its additive homomorphism under the homomorphic encryption algorithm Paillier, then the computation result is as follows:

for

And

these two parts, using additive homomorphism, are converted to the following equation:

II: performing continuous multiplication operation;

for the

This section, using additive homomorphism, converts it into the following equation:

the computing server CS will

Conversion is carried out, and the conversion formula is as follows:

having plaintext, ciphertext templates T in the computing server CS _A ，

And on the premise of the encrypted candidate template set, the computing server CS checks the ciphertext template

And the Euclidean distance of each candidate template, the minimum Euclidean distance in the result

Whether a set threshold T is met; if the value is smaller than the set threshold value T, the computing server CS considers that the authentication is passed; otherwise, the computing server CS considers the authentication as failed.

7. The method of claim 6, wherein the feature update phase comprises:

the system supports three update operations: adding, deleting and modifying, namely registering a new user, canceling an existing user and updating an existing user key and a feature template based on a revocable template module, the specific process is as follows:

(1) New user U _ADD Uploading own biological characteristic data through an extractor, processing the biological characteristic data by the extractor, and respectively generating encrypted registration vectors

And encrypted feature templates

The extractor will

To the calculation server CS, which sends it to the database DB, which will

Adding the new user into the stored encryption index to complete the registration of the new user;

(2) The existing user revocation process requires three operations; first, the extractor collects and extracts the users U to be revoked _DEL The biological characteristic of (a); in addition, the extractor utilizes the registration key

Generating matching index entries

The extractor then indexes the entry

Sending the data to a computing server CS, and sending the data to a database DB by the computing server CS; finally, the database DB deletes the index entry on the index it stores

And matched encryption template

(3) When the user U _i When the biological template is damaged or stolen, the user U _i Re-inputting the biological characteristics based on the template module which can be cancelled; first, an extractor extracts a user biometric feature and obtains a registration index item

And an encryption template

In addition, the extractor generates new enrollment and template keys

And

then, the extractor will

And

sending the data to a computing server CS, and sending the data to a database DB by the computing server CS; finally, the database DB utilizes the index entry

Dematching encryption index I with user U _i Related indexing item

Deleting an index entry

And associated cryptographic templates

And will be

And

inserted into the encryption index I.

8. A storage medium for use in a method according to any of claims 1-7, wherein the program storage medium is stored for enabling an electronic device to execute via a computer program by receiving user input.

9. A lightweight biometric authentication system for use in the method of claims 1-8, wherein the lightweight biometric authentication system comprises:

the extractor is used for extracting biological characteristics, generating a template module which can be cancelled and encrypting a template;

a trusted center for generating a key;

a calculation server for calculating a euclidean distance of the encrypted biometric feature;

and the database is used for storing the indexes.

10. The system of claim 9, wherein the lightweight biometric authentication system is mounted on a terminal, and the terminal is an internet of things terminal.

Images