Single-factor revocable face template protection method and system

Abstract

The invention relates to the technical field of biological feature recognition, in particular to a single-factor revocable face template protection method and system, and the method comprises the steps: generating a registered user face Hash code through a face Hash code generation algorithm, carrying out the XOR operation of the registered user face Hash code and a random binary bit string to generate a user specific secret key, and storing the user specific secret key in a database; performing one-time permutation hash operation on the random binary bit string to obtain a registered user face template, and storing the user specific key and the corresponding registered user face template to a face template database; a target face hash code of a to-be-authenticated user is generated through a face hash code generation algorithm, XOR operation is performed on the target face hash code and a user specific key in a face template database to obtain a corresponding middle bit string, and one-time permutation hash operation is performed on the middle bit string to obtain a query face template of the to-be-authenticated user. And judging whether the to-be-authenticated user passes authentication or not through similarity matching. According to the invention, the problem of potential safety hazards in a two-factor revocable face template protection method can be effectively avoided.

Description

Technical Field

[0001] The present invention relates to the field of biometric recognition technology, and in particular to a single-factor revocable face template protection method and system. Background Art

[0002] With the widespread use of face recognition systems, there is a risk that face templates cannot be reissued once lost. To address this issue, researchers have proposed a revocable face template protection method. When a user's face template is leaked or attacked, the revocable face template protection method can use new parameters to generate a new face template for the user, making the face template revocable. Among them, in the two-factor revocable face template protection method, external factors (user-specific parameters, usually in the form of tokens) and facial features are required as input to jointly participate in the face template generation process. However, the use of external factors brings inconveniences to users, such as saving tokens or remembering passwords. At the same time, external factors are at risk of being stolen, lost, or forgotten, posing a serious threat to users' biometric privacy. Summary of the Invention

[0003] To this end, the present invention provides a single-factor revocable face template protection method and system to solve the security risks existing in the existing two-factor revocable face template protection method.

[0004] According to the design scheme provided by the present invention, on the one hand, a single-factor revocable face template protection method is provided, comprising:

[0005] Collect a facial image of a registered user, generate a facial hash code for the registered user using a face hash code generation algorithm, perform an XOR operation on the facial hash code of the registered user and a random binary bit string to generate a user-specific key, perform a permutation hash on the random binary bit string to obtain a facial template of the registered user, and store the user-specific key and the corresponding facial template of the registered user in a facial template database. The face hash code generation algorithm generates a facial hash code through facial image feature extraction, a permutation hash algorithm, and a hash code copy and extension operation.

[0006] Collect the facial image of the user to be authenticated, generate the target facial hash code of the user to be authenticated through the face hash code generation algorithm, perform an XOR operation on the target facial hash code and the user-specific key in the face template database to obtain the corresponding intermediate bit string, perform a permutation hash on the intermediate bit string to obtain the query facial template of the user to be authenticated, and perform similarity matching between the query facial template and the registered user facial template corresponding to the user-specific key to determine whether the facial image of the user to be authenticated is a facial image of a registered user.

[0007] As a single-factor revocable face template protection method of the present invention, further, a face hash code generation algorithm is used to generate a registered user's face hash code, including:

[0008] Extract facial image features and form facial feature vectors;

[0009] For the facial feature vector, a face hash code is generated using a single permutation hash algorithm, wherein the single permutation hash algorithm generates a hash code through a single random permutation operation;

[0010] The face hash code is copied and expanded to obtain a high-dimensional face hash code.

[0011] As the single-factor revocable face template protection method of the present invention, further extracting face image features includes:

[0012] Align and crop the collected facial images to obtain facial images of fixed size;

[0013] A fixed-size face image is input into a deep neural network model, and the deep neural network model is used to extract facial features of the face image and generate a facial feature vector. The deep neural network model is obtained by training the face recognition network ArcFace using a face image dataset based on the additive angular margin loss function. The additive angular margin loss function guides network training optimization by maximizing the inter-class distance and minimizing the intra-class distance in the feature space.

[0014] As a single-factor revocable face template protection method of the present invention, further, a one-time permutation hash algorithm includes:

[0015] Group two adjacent elements in the facial feature vector into a group and mark the index value of each group of elements;

[0016] The index value of the smallest element in each group is used as the hash code.

[0017] As the single-factor revocable face template protection method of the present invention, further, the random binary bit string is generated by a random binary number generator.

[0018] As a single-factor revocable face template protection method of the present invention, further, a permutation hash operation is performed on the random binary bit string to obtain a registered face template, including:

[0019] Group two adjacent elements in the random binary bit string into a group, and mark the two elements in each group with index values;

[0020] The index value of the smallest element in each group is taken as the hash code, and the required registered face template is obtained according to the hash code generated by the random binary bit string.

[0021] As a single-factor revocable face template protection method of the present invention, further, similarity matching is performed between the query face template and the registered face template corresponding to the user-specific key, including:

[0022] The similarity between the query face template and the registered face templates in the face template database is calculated by calculating the total number of hash code values at the same position between the face templates based on the Jaccard similarity;

[0023] Determine whether the query face template matches the registered face template in the face template database based on the similarity.

[0024] On the other hand, the present invention also provides a single-factor revocable face template protection system, comprising: a registration module and a recognition module, wherein:

[0025] A registration module is configured to collect a facial image of a registered user, generate a facial hash code for the registered user using a facial hash code generation algorithm, perform an XOR operation on the facial hash code of the registered user and a random binary bit string to generate a user-specific key, perform a permutation hash operation on the random binary bit string to obtain a facial template of the registered user, and store the user-specific key and the corresponding facial template of the registered user in a facial template database. The facial hash code generation algorithm generates a facial hash code by extracting facial image features, performing a permutation hash algorithm, and performing a hash code copy and extension operation.

[0026] The recognition module is used to collect the face image of the user to be authenticated, generate the target face hash code of the user to be authenticated through the face hash code generation algorithm, perform an XOR operation on the target face hash code and the user-specific key in the face template database to obtain the corresponding intermediate bit string, perform a permutation hash operation on the intermediate bit string to obtain the query face template of the user to be authenticated, and perform similarity matching between the query face template and the registered face template corresponding to the user-specific key to determine whether the face image of the user to be authenticated is a face image of a registered user.

[0027] Beneficial effects of the present invention:

[0028] The present invention only uses facial features as the only input factor. When the user authenticates, the user's facial features as a single factor are involved in the generation process of the query template, which can effectively avoid the occurrence of token theft scenarios (that is, the attacker uses the leaked real token to launch a zero-effort false acceptance attack), and can avoid the situation where the attacker recovers the original facial features from the face template when the external factors are leaked. It can effectively protect the user's biometric privacy and has good application prospects in the field of face recognition identity authentication and authorization. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1This is a schematic diagram of the single-factor revocable face template protection process in the embodiment;

[0030] Figure 2 This is a schematic diagram of the overall framework of the single-factor revocable face template protection method in the embodiment;

[0031] Figure 3 This is an illustration of generating a registration face template and a query face template when n=2 in the embodiment. DETAILED DESCRIPTION

[0032] In order to make the objectives, technical solutions and advantages of the present invention clearer and more understandable, the present invention is further described in detail below with reference to the accompanying drawings and technical solutions.

[0033] For the dual-factor revocable face template protection method, when a user authenticates, both the external factor (token) and the facial feature are required as input to generate the user's query template. However, the user-specific external factors are usually stored in an external memory (e.g., a personal memory or physical hardware), so there are security threats such as leakage and loss of external factors. Therefore, in the embodiments of the present invention, Figure 1 As shown, a single-factor revocable face template protection method is provided, which specifically includes the following contents:

[0034] S101. Collect a facial image of a registered user, generate a facial hash code for the registered user using a facial hash code generation algorithm, perform an XOR operation on the facial hash code of the registered user and a random binary bit string to generate a user-specific key, perform a permutation hash operation on the random binary bit string to obtain a facial template of the registered user, and store the user-specific key and the corresponding facial template of the registered user in a facial template database. The facial hash code generation algorithm generates a facial hash code by extracting facial image features, performing a permutation hash algorithm, and performing a hash code copy and extension operation.

[0035] S102. Collect a facial image of the user to be authenticated, generate a target facial hash code of the user to be authenticated through a facial hash code generation algorithm, perform an XOR operation on the target facial hash code and the user-specific key in the facial template database to obtain a corresponding intermediate bit string, perform a permutation hash operation on the intermediate bit string to obtain a query facial template of the user to be authenticated, perform a similarity match between the query facial template and the registered facial template corresponding to the user-specific key to determine whether the facial image of the user to be authenticated is a facial image of a registered user.

[0036] Among them, the face hash code of the registered user is generated by the face hash code generation algorithm, which can be designed to include:

[0037] Extract facial image features and form facial feature vectors;

[0038] For the facial feature vector, a face hash code is generated using a single permutation hash algorithm, wherein the single permutation hash algorithm generates a hash code through a random permutation operation;

[0039] The face hash code is copied and expanded to obtain a high-dimensional face hash code.

[0040] like Figure 2 In the algorithm architecture shown, in the registration phase, the face hash code C is first generated based on a permutation hash algorithm, and then the face hash code C is copied and expanded to obtain Then the binary bit string r generated by the random binary number generator is combined with the copied and expanded face hash code Perform an XOR operation to generate a user-specific key K, and finally generate a hash code of the random binary bit string r based on a one-time permutation hash algorithm as the registration template T. In the authentication stage, the face hash code C' is also first generated based on a one-time permutation hash algorithm, and then the face hash code is copied and expanded to obtain Then copy the expanded face hash code The random binary bit string r' is recovered by performing an XOR operation with the generated user-specific key K. Finally, a hash code of the recovered random binary bit string r' is generated using a single-permutation hash algorithm as the query template T'. By calculating the similarity between the registration template T and the query template T', it can be determined whether the user has passed system authentication.

[0041] Specifically, extracting facial image features may include:

[0042] Align and crop the collected facial images to obtain facial images of fixed size;

[0043] A fixed-size face image is input into a deep neural network model, and the deep neural network model is used to extract facial features of the face image and generate a facial feature vector. The deep neural network model is obtained by training the face recognition network ArcFace using a face image dataset based on the additive angular margin loss function. The additive angular margin loss function guides network training optimization by maximizing the inter-class distance and minimizing the intra-class distance in the feature space.

[0044] The face image is aligned and cropped to 112×112 size by MTCNN; the ArcFace model pre-trained on the MS-Celeb-1M dataset is used to extract the 512-dimensional face feature vector ArcFace is a deep learning-based face recognition model that uses a novel additive angular margin loss function to improve the model's discriminative ability by increasing the angular margin. The core idea of ArcFace is to maximize the inter-class distance and minimize the intra-class distance in feature space, thereby learning more discriminative facial features.

[0045] Specifically, the single-permutation hash algorithm groups two adjacent elements in the facial feature vector into a group and marks the index value of each group of elements; the index value of the minimum element in each group is used as the hash code.

[0046] That is, the two adjacent elements in the face feature vector x are grouped together, and the index values of the two elements can be recorded as 0 and 1 respectively. Since the dimension of the face feature vector x is 512, 256 groups can be obtained. Then, based on the idea of a single permutation hash algorithm, the index value of the smallest element in each group is taken as the hash code in Represents the minimum element index value in each group.

[0047] In order to ensure the security and recognition accuracy of the face template, the face hash code C is copied and expanded n times to obtain a high-dimensional face hash code Its dimension is 256×n. The specific method is to concatenate n 256-dimensional face hash codes C together end to end.

[0048] Specifically, a permutation hash operation is performed on the random binary bit string to obtain a registered face template, which can be designed to include:

[0049] Group two adjacent elements in the random binary bit string into a group, and mark the two elements in each group with index values;

[0050] The index value of the smallest element in each group is taken as the hash code, and the required registered face template is obtained according to the hash code generated by the random binary bit string.

[0051] The expanded face hash code will be copied The user-specific key K can be generated by performing an XOR operation with the random binary bit string r generated by the random binary number generator, that is,

[0052]

[0053] in The dimensions of r and K are the same, both 256×n.

[0054] Group two adjacent elements in the random binary bit string r into a group, and record the index values of the two elements as 0 and 1 respectively. Since the dimension of the random binary bit string r is 256×n, 128×n groups can be obtained. Then, based on the idea of a single permutation hash algorithm, the index value of the smallest element in each group is taken as the registered face template in Represents the minimum element index value in each group.

[0055] Similarly, in the authentication phase, the deep neural network model ArcFace is used to extract the facial feature vector. Since the dimension of the facial feature vector x' is 512, 256 groups can be obtained. Then, based on the idea of a single permutation hash algorithm, the index value of the smallest element in each group is taken as the hash code in Represents the minimum element index value in each group. In order to ensure the security and recognition accuracy of the face template, the face hash code C' is copied and expanded n times to obtain a high-dimensional face hash code Its dimension is 256×n. The specific method is to concatenate n 256-dimensional face hash codes C' together. The random binary bit string r' can be restored by performing an XOR operation with the generated user-specific key K, that is,

[0056]

[0057] in K and r' have the same dimension, both 256×n.

[0058] The two adjacent elements in the recovered random binary bit string r' are grouped together, and the index values of the two elements are recorded as 0 and 1 respectively. Since the dimension of the random binary bit string r' is 256×n, 128×n groups can be obtained. Then, take the index value of the minimum element in each group as the query face template in Represents the minimum element index value in each group.

[0059] Based on the Jaccard similarity, the total number of identical hash code values at the same position between the face templates is calculated to calculate the similarity between the query face template and the registered user face template in the face template database; based on the similarity, it is determined whether the query face template matches the registered user face template in the face template database.

[0060] like Figure 2 and 3 As shown, by calculating the similarity between the registration template and the query template generated by the user during the registration and authentication stages, it can be determined whether the user can pass the authentication. The specific calculation formula for calculating the similarity between the registration template T and the query template T' based on the Jaccard similarity can be expressed as follows:

[0061]

[0062] Among them, 128n represents the length of T and T', XNOR(T,T') represents the exclusive OR operation, when the hash code values of the corresponding positions of T and T' are the same, it outputs 1, and when they are different, it outputs 0, so The total number of hash code values at the same position in T and T' is calculated. S(T,T'))∈[0,1], the closer its value is to 1, the higher the similarity between T and T', and vice versa.

[0063] Furthermore, based on the above method, an embodiment of the present invention also provides a single-factor revocable face template protection system, comprising: a registration module and a recognition module, wherein:

[0064] A registration module, configured to collect a facial image of a registered user, generate a facial hash code for the registered user using a facial hash code generation algorithm, perform an XOR operation on the facial hash code of the registered user and a random binary bit string to generate a user-specific key, perform a permutation hash operation on the random binary bit string to obtain a facial template of the registered user, and store the user-specific key and the corresponding facial template of the registered user in a facial template database. The facial hash code generation algorithm generates the facial hash code through facial image feature extraction, a permutation hash algorithm, and a hash code copy and extension operation.

[0065] The recognition module is used to collect the face image of the user to be authenticated, generate the target face hash code of the user to be authenticated through the face hash code generation algorithm, perform an XOR operation on the target face hash code and the user-specific key in the face template database to obtain the corresponding intermediate bit string, perform a permutation hash operation on the intermediate bit string to obtain the query face template of the user to be authenticated, and perform similarity matching between the query face template and the registered face template corresponding to the user-specific key to determine whether the face image of the user to be authenticated is a face image of a registered user.

[0066] Unless otherwise specifically stated, the relative steps, numerical expressions and values of the components and steps set forth in these embodiments do not limit the scope of the present invention.

[0067] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. Reference can be made to the common and similar parts between the various embodiments. For the systems disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple, and the relevant parts can be referred to the method description.

[0068] The units and method steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the components and steps of each example have been generally described in terms of function in the above description. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. A person of ordinary skill in the art may use different methods to implement the described functions for each specific application, but such implementation is not considered to be beyond the scope of the present invention.

[0069] Those skilled in the art will appreciate that all or part of the steps in the above method can be performed by a program instructing related hardware. The program can be stored in a computer-readable storage medium, such as a read-only memory, a magnetic disk, or an optical disk. Alternatively, all or part of the steps in the above embodiment can be implemented using one or more integrated circuits. Accordingly, each module / unit in the above embodiment can be implemented in the form of hardware or software functional modules. The present invention is not limited to any specific combination of hardware and software.

[0070] Finally, it should be noted that the above-described embodiments are only specific implementation methods of the present invention, which are used to illustrate the technical solutions of the present invention, rather than to limit them. The scope of protection of the present invention is not limited thereto. Although the present invention has been described in detail with reference to the above-described embodiments, those skilled in the art should understand that any person skilled in the art can modify or easily conceive of changes to the technical solutions described in the above-described embodiments within the technical scope disclosed by the present invention, or replace some of the technical features therein with equivalents. Such modifications, changes, or replacements do not deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should be included in the scope of protection of the present invention. Therefore, the scope of protection of the present invention shall be subject to the scope of protection of the claims.

Claims

1. A single-factor revocable face template protection method, characterized in that: Include: Collect a facial image of a registered user, generate a facial hash code for the registered user using a face hash code generation algorithm, perform an XOR operation on the facial hash code of the registered user and a random binary bit string to generate a user-specific key, perform a permutation hash operation on the random binary bit string to obtain a facial template of the registered user, and store the user-specific key and the corresponding facial template of the registered user in a facial template database. The face hash code generation algorithm generates a facial hash code through facial image feature extraction, a permutation hash algorithm, and a hash code copy and extension operation. Collect the facial image of the user to be authenticated, generate the target facial hash code of the user to be authenticated through the face hash code generation algorithm, perform an XOR operation on the target facial hash code and the user-specific key in the face template database to obtain the corresponding intermediate bit string, perform a permutation hash operation on the intermediate bit string to obtain the query facial template of the user to be authenticated, perform similarity matching between the query facial template and the registered facial template corresponding to the user-specific key to determine whether the facial image of the user to be authenticated is the facial image of a registered user.

2. The single-factor revocable face template protection method according to claim 1, characterized in that: Generate a registered user's face hash code through a face hash code generation algorithm, including: Extract facial image features and form facial feature vectors; For the facial feature vector, a face hash code is generated using a single permutation hash algorithm, wherein the single permutation hash algorithm generates a hash code through a random permutation operation; The face hash code is copied and expanded to obtain a high-dimensional face hash code.

3. The single-factor revocable face template protection method according to claim 1 or 2, characterized in that: Extract facial image features, including: Align and crop the collected facial images to obtain facial images of fixed size; A fixed-size face image is input into a deep neural network model, and the deep neural network model is used to extract facial features of the face image and generate a facial feature vector. The deep neural network model is obtained by training the face recognition network ArcFace using a face image dataset based on the additive angular margin loss function. The additive angular margin loss function guides network training optimization by maximizing the inter-class distance and minimizing the intra-class distance in the feature space.

4. The single-factor revocable face template protection method according to claim 1 or 2, characterized in that: A permutation hash algorithm, including: Group two adjacent elements in the facial feature vector into a group and mark the index value of each group of elements; The index value of the smallest element in each group is used as the hash code.

5. The single-factor revocable face template protection method according to claim 1, characterized in that: The random binary bit string is generated by a random binary number generator.

6. The single-factor revocable face template protection method according to claim 1, characterized in that: Perform a permutation hash operation on the random binary bit string to obtain a face template, which contains: Group two adjacent elements in the bit string into a group, and mark the two elements in each group with index values; The index value of the smallest element in each group is taken as the hash code, and the required face template is obtained based on the hash code generated by the bit string.

7. The single-factor revocable face template protection method according to claim 1, characterized in that: Perform similarity matching between the query face template and the registered user face template corresponding to the user-specific key, including: The similarity between the query face template and the registered user face template in the face template database is calculated based on the Jaccard similarity by calculating the total number of hash code values at the same position between the face templates; Determine whether the query face template matches the registered user face template in the face template database based on the similarity.

8. A single-factor revocable face template protection system, characterized in that: The system comprises a registration module and an identification module, wherein the registration module is used to collect a facial image of a registered user, generate a facial hash code of the registered user through a facial hash code generation algorithm, perform an XOR operation on the facial hash code of the registered user and a random binary bit string to generate a user-specific key, perform a permutation hash operation on the random binary bit string to obtain a facial template of the registered user, store the user-specific key and the corresponding facial template of the registered user in a facial template database, and the facial hash code generation algorithm generates a facial hash code through facial image feature extraction, a permutation hash algorithm, and a hash code copy and extension operation; The recognition module is used to collect the face image of the user to be authenticated, generate the target face hash code of the user to be authenticated through the face hash code generation algorithm, perform an XOR operation on the target face hash code and the user-specific key in the face template database to obtain the corresponding intermediate bit string, perform a permutation hash operation on the intermediate bit string to obtain the query face template of the user to be authenticated, and perform similarity matching between the query face template and the registered face template corresponding to the user-specific key to determine whether the face image of the user to be authenticated is a face image of a registered user.

9. An electronic device, characterized in that: include: at least one processor, and a memory coupled to the at least one processor; The memory stores a computer program, and the computer program can be executed by the at least one processor to implement the method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that The computer program stored in the computer-readable storage medium can implement the method according to any one of claims 1 to 7 when the computer program is executed.

Images