Palm vein feature template generation method, device and equipment for resisting similarity attacks

Abstract

Embodiments of the present invention provide a method, apparatus, and device for generating a palm vein feature template resistant to similarity attacks. The method includes obtaining original palm vein data and obtaining an initial palm vein feature vector according to the original palm vein data; performing random projection on the initial palm vein feature vector to obtain a randomized palm vein feature vector; using multi-dimensional spectral hashing on the randomized palm vein feature vector to generate a first palm vein feature template; performing a non-linear transformation on the first palm vein feature template using a non-linear activation function to obtain a second palm vein feature template; and performing binary conversion on the second palm vein feature template to obtain a third palm vein feature template. In this way, similarity attacks can be resisted while maintaining performance loss, with higher security; by using the spectral segmentation technology, the similarity structure of the palm vein data is preserved and concentrated, reducing the constraint of high security on performance, and achieving a balance between security and performance relative to existing palm vein template protection schemes.

Description

Technical Field

[0001] The present invention generally relates to the technical field of feature extraction and recognition methods, and more specifically, to a method, apparatus, and device for generating a palm vein feature template resistant to similarity attacks. Background Art

[0002] As a biometric feature inside the human body, palm veins have high concealment and have attracted a large amount of biometric recognition research in recent years. At the same time, due to the leakage and security risks of palm vein data, a large number of research on palm vein feature template protection methods have also emerged. The international standard ISO / IEC 30136:2018 "Information technology—Performance testing of biometric template protection schemes" stipulates that an ideal biometric template protection scheme should not cause serious performance losses while ensuring the security rate of the biometric recognition system. Based on this, in order to meet the recognition accuracy requirements of the ideal biometric protection scheme, the existing palm vein template protection-related methods often make the similarity of the final template distribution space consistent with the similarity of the original biometric sample space. By maintaining this similarity in distribution, the loss of recognition rate is reduced. However, this similarity-preserving property provides an optimization direction for similarity attacks, which are a type of reconstruction attack. As a result, in recent years, reconstruction attacks that use similarity preservation to reconstruct input biometric data have emerged in an endless stream, and these reconstructed biometric data can pass the authentication of the biometric recognition system. This similarity attack poses a serious security threat to the existing biometric recognition systems. In the face of this attack method, the research on palm vein template protection schemes is still insufficient. Summary of the Invention

[0003] According to an embodiment of the present invention, a palm vein feature template generation scheme resistant to similarity attacks is provided. This scheme can resist similarity attacks while maintaining performance losses, and has higher security; by using the atlas segmentation technology, the similarity structure of the palm vein data is preserved and concentrated, reducing the restriction of high security on performance, and achieving a balance between security and performance relative to the existing palm vein template protection schemes.

[0004] In a first aspect of the present invention, a method for generating a palm vein feature template resistant to similarity attacks is provided. The method includes:

[0005] Obtain original palm vein data, and obtain an initial palm vein feature vector according to the original palm vein data;

[0006] Perform random projection on the initial palm vein feature vector to obtain a randomized palm vein feature vector;

[0007] Use multi-dimensional spectral hashing on the randomized palm vein feature vector to generate a first palm vein feature template;

[0008] Perform a non-linear transformation on the first palm vein feature template using a non-linear activation function to obtain a second palm vein feature template;

[0009] Perform binary conversion on the second palm vein feature template to obtain a third palm vein feature template.

[0010] Further, obtaining the initial palm vein feature vector according to the original palm vein data includes:

[0011] Map the original palm vein data to the trained projection matrix to obtain the initial palm vein feature vector;

[0012] The trained projection matrix is obtained by iterative calculation using the SDSPCA-NPE method.

[0013] Further, performing random projection on the initial palm vein feature vector includes:

[0014] Generate a pseudo-random matrix using a predefined key;

[0015] Convert the pseudo-random matrix into a standard orthogonal matrix using the Schmidt orthogonalization method;

[0016] Input the initial palm vein feature vector into the standard orthogonal matrix to obtain a randomized palm vein feature vector.

[0017] Further, inputting the initial palm vein feature vector into the standard orthogonal matrix to obtain a randomized palm vein feature vector includes:

[0018] X3 = X2 × R1, X3 ∈ n

[0019] where X3 represents the randomized palm vein feature vector; X2 represents the initial palm vein feature vector; R1 represents the standard orthogonalized pseudo-random matrix; represents the set of real numbers; n represents the dimension of the vector.

[0020] Further, using multi-dimensional spectral hashing on the randomized palm vein feature vector to generate a first palm vein feature template includes:

[0021] First, calculate for each column in the randomized palm vein feature vector, and calculate the one-dimensional feature function φ ij (x(i));

[0022]

[0023] Wherein, i represents the serial number of the first column in the randomized palm vein feature vector; j represents the serial numbers of columns other than the first column in the randomized palm vein feature vector; a i represents the minimum value in the column data of the randomized palm vein feature vector; b i represents the maximum value in the column data of the randomized palm vein feature vector; x(i) is an arbitrary one-dimensional real feature corresponding to the i-th coordinate of the randomized palm vein feature vector x;

[0024] Secondly, calculate the eigenvalue λ corresponding to each column data ij ;

[0025]

[0026] Wherein, e represents the natural constant; δ 2 represents the variance of each column data;

[0027] Finally, select the column elements corresponding to several largest eigenvalues, rearrange the column elements in the order of the eigenvalue magnitudes to form a set A, and generate a first palm vein feature template.

[0028] Further, the non-linearly transforming the first palm vein feature template by using a non-linear activation function to obtain a second palm vein feature template includes:

[0029] y2 = q(y1) = q(φ ij (B))

[0030]

[0031] Wherein, q(x) represents the non-linear activation function; tanh represents the hyperbolic tangent function; φ ij (B) represents a one-dimensional feature function; α represents a label information weight parameter, α = 0.4; y1 represents the first palm vein feature template; y2 represents the second palm vein feature template; B represents a feature matrix obtained by rearranging the column elements in the randomized palm vein feature vector X3 that satisfy the index of the set A.

[0032] Further, the binary converting the second palm vein feature template to obtain a third palm vein feature template includes:

[0033] y3 = sign(y2) = sign(q(φ ij (B)))

[0034] Wherein, y3 represents the third palm vein feature template, and sign represents the binary sign function.

[0035] In the second aspect of the present invention, a palm vein feature template generating device for resisting similarity attacks is provided. The device includes:

[0036] An acquisition module, configured to acquire original palm vein data and obtain an initial palm vein feature vector according to the original palm vein data;

[0037] A random projection module, configured to perform random projection on the initial palm vein feature vector to obtain a randomized palm vein feature vector;

[0038] A first template generation module, configured to use multi-dimensional spectral hashing on the randomized palm vein feature vector to generate a first palm vein feature template;

[0039] A second template generation module, configured to perform non-linear transformation on the first palm vein feature template by using a non-linear activation function to obtain a second palm vein feature template;

[0040] A third template generation module, configured to perform binary conversion on the second palm vein feature template to obtain a third palm vein feature template.

[0041] In a third aspect of the present invention, there is provided an electronic device. The electronic device includes at least one processor; and a memory communicatively connected to the at least one processor; the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the method of the first aspect of the present invention.

[0042] In a fourth aspect of the present invention, there is provided a non-transitory computer-readable storage medium storing computer instructions, and the computer instructions are used to cause the computer to execute the method of the first aspect of the present invention.

[0043] It should be understood that the content described in the summary of the invention is not intended to limit the key or important features of the embodiments of the present invention, nor is it used to limit the scope of the present invention. Other features of the present invention will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In combination with the accompanying drawings and with reference to the following detailed description, the above and other features, advantages and aspects of the embodiments of the present invention will become more obvious. In the drawings, the same or similar reference numerals denote the same or similar elements, where:

[0045] Figure 1 shows a flowchart of a method for generating a palm vein feature template resistant to similarity attacks according to an embodiment of the present invention;

[0046] Figure 2 shows a block diagram of a random projection template according to an embodiment of the present invention;

[0047] Figure 3 shows a block diagram of using multi-dimensional spectral hashing on a randomized palm vein feature vector according to an embodiment of the present invention;

[0048] Figure 4 A block diagram of a palm vein feature template generation device for resisting similarity attacks according to an embodiment of the present invention is shown;

[0049] Figure 5 A block diagram of an exemplary electronic device capable of implementing the embodiments of the present invention is shown;

[0050] Wherein, 500 is an electronic device, 501 is a computing unit, 502 is a ROM, 503 is a RAM, 504 is a bus, 505 is an I / O interface, 506 is an input unit, 507 is an output unit, 508 is a storage unit, and 509 is a communication unit. Detailed implementation manners

[0051] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Apparently, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without making creative efforts shall fall within the protection scope of the present invention.

[0052] In addition, the term "and / or" in this article is merely a description of the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B may represent: A exists alone, A and B exist simultaneously, and B exists alone. In addition, the character " / " in this article generally represents an "or" relationship between the associated objects before and after.

[0053] In the present invention, after using Multi-Dimension Spectral Hashing (MDSH) to generate a palm vein template with acceptable performance attenuation, while ensuring the security and accuracy of the palm vein features, by performing a non-linear periodic transformation on the palm vein feature template, the number of palm vein feature templates with large differences is reduced. The inter-class distance of the palm vein feature templates is made as similar as possible, and the requirements of mutual information can be met within a certain range, so that the feature templates generated by the proposed method have the ability to resist similarity attacks.

[0054] Figure 1 A flowchart of a palm vein feature template generation method for resisting similarity attacks according to an embodiment of the present invention is shown.

[0055] The method includes:

[0056] S101. Obtain original palm vein data, and obtain an initial palm vein feature vector according to the original palm vein data.

[0057] In this embodiment, mapping the original palm vein data to the trained projection matrix to obtain an initial palm vein feature vector includes:

[0058] X2 = X1 * W

[0059] where W represents the trained projection matrix; X1 represents the original palm vein data; and X2 represents the initial palm vein feature vector.

[0060] In this embodiment, the training process of the projection matrix model includes: selecting a training set from a palm vein image library, and the trained projection matrix is obtained by iterative calculation using the SDSPCA-NPE (supervised discriminative sparse principal component analysis algorithm with a pre-served neighborhood structure) method.

[0061] Specifically, the palm vein image library includes the Tongji University image library, the Institute of Automation, Chinese Academy of Sciences image library, the Hong Kong Polytechnic University image library, and a self-built image library. SDSPCA-NPE is a supervised discriminative sparse principal component analysis algorithm that preserves the neighborhood structure and includes:

[0062] Step 1: Construct an initial palm vein feature data matrix:

[0063] X = [x1,…,x m T ∈ m×d

[0064] where X represents the initial palm vein feature data matrix; m represents the feature dimension; d represents the number of samples; and represents the set of real numbers.

[0065] Step 2: Iteratively solve Q using the following formula:

[0066]

[0067] where Tr represents the trace of a matrix; Q represents the auxiliary matrix of matrix W; Q T represents the transpose of the auxiliary matrix of matrix W; X T represents the transpose matrix of the initial palm vein feature data matrix X; Y represents the label matrix of the initial palm vein feature data matrix X; Y T represents the transpose matrix of the label matrix Y; D represents an auxiliary matrix to prevent overfitting; M represents the local information matrix; α represents the label information weight parameter; β represents the regularization parameter; and δ represents the local information weight.

[0068] The label matrix of the dataset X: ​

[0069] Y = [y1, …, y m T ∈ m×c

[0070] where c represents the number of classes in the training data.

[0071] Y is specifically constructed as follows:

[0072]

[0073] where c j ∈ {1, …, c} represents the class label; s represents the number of rows of the matrix; t represents the number of columns of the matrix; if represents the condition.

[0074] D ∈ m×m is a diagonal matrix. When s = t, the diagonal elements are:

[0075]

[0076] where k represents the specific cumulative quantity; represents (the square of the element in the s-th row and t-th column of Q); ∈ represents a small positive constant to avoid division by zero.

[0077] The optimal Q consists of the first L eigenvectors of Z = -XX T - αYY T + βD + δXX T MXX T where Z represents the entire constraint function. Initialize Q, calculate D according to the given formula, and obtain Z. With Z and D known, update the Q value through this iterative process until the optimal Q is reached.

[0078] Step 3: Calculate the trained projection matrix according to the following formula:

[0079] W = X T * Q

[0080] where W represents the trained projection matrix.

[0081] By introducing local structure information, SDSPCA - NPE enables the final projection matrix W to maintain certain neighborhood relationships of the original palm vein data, solving the limitation of only considering global and class information in supervised learning. This improves the tolerance to outliers with non - uniform distribution and reduces the within - class distance. It effectively reduces the impact of illumination changes in palm vein images and improves the problem of high inter - class similarity. Therefore, the classification performance of the final palm vein feature vectors is improved, laying a good foundation for the performance of subsequent template protection methods.

[0082] ​S102. Perform random projection on the initial palm vein feature vector to obtain a randomized palm vein feature vector.

[0083] As Figure 2 shown in the block diagram of the random projection template according to an embodiment of the present invention, the specific steps of random projection are as follows:

[0084] S201. Generate a pseudo-random matrix using a predefined key.

[0085] Specifically, given a key K, the generation formula is:

[0086] R = K * rand(E, F)

[0087] where R represents the pseudo-random matrix; K represents the key; E represents the number of rows of the pseudo-random matrix, which needs to be the same as the number of columns of the training matrix; F represents the number of columns of the pseudo-random matrix.

[0088] S202. Use the Gram-Schmidt orthogonalization method to convert the pseudo-random matrix into a standard orthogonal matrix.

[0089] Specifically, the Gram-Schmidt process is a general concept in mathematics, and the specific process is as follows:

[0090] The pseudo-random matrix R = {a1, a2,..., a n}, where R represents the pseudo-random matrix and a i represents the elements of the pseudo-random matrix.

[0091] The formula for converting the pseudo-random matrix into a standard orthogonal matrix is as follows:

[0092] β1 = a1

[0093]

[0094] ……………

[0095]

[0096] where β1 represents the first element to be orthogonalized; β2 represents the second element to be orthogonalized; β3 represents the third element to be orthogonalized; β g represents the g-th element to be orthogonalized; a1 represents the first element of the pseudo-random matrix; a2 represents the second element of the pseudo-random matrix; a3 represents the third element of the pseudo-random matrix; g represents the index of the elements of the pseudo-random matrix; α g represents the g-th element of the pseudo-random matrix; γ g represents the elements of the standardized orthogonal pseudo-random matrix.

[0097] Through the above transformation, we can get the standard orthogonal matrix R1 = {γ1,γ2,…,γ g}, where R1 represents the standard orthogonalized pseudo-random matrix.

[0098] S203, inputting the initial palm vein feature vector into the standard orthogonal matrix to obtain a randomized palm vein feature vector, including:

[0099] X3=X2×R1,X3∈ n

[0100] Among them, X3 represents the randomized palm vein feature vector; X2 represents the initial palm vein feature vector; R1 represents the standard orthogonalized pseudo-random matrix; represents the real number set; n represents the dimension of the vector.

[0101] Through random projection, once the palm vein feature template is cracked, a new protected template can be quickly generated by executing the proposed method using different random projection keys to avoid further leakage risks.

[0102] S103: Use multi-dimensional spectral hashing on the randomized palm vein feature vector to generate a first palm vein feature template.

[0103] Specifically, multidimensional spectral hashing (MDSH) is a type of locality sensitive hashing (LSH), which was originally proposed for image retrieval applications based on spectral graph partitioning. Multidimensional spectral hashing uses an affinity matrix to indicate the similarity between given data. p and x q The affinity between them is defined as Where σ represents the distance weight parameter, AFF represents affinity, x represents the data point, p represents the range of p-class points in space, and q represents the range of q-class points in space. To learn binary encoding, the typical cost function is based on the data point x p and x q The Hamming distance between binary codes of ||y p -y q || 2 Given, where y p represents a binary vector of length o, y q represents the q-category binary vector. p The elements of y consist only of 1 or -1, so the Hamming distance can be defined as a simple dot product function y p T y q , because ||y p -y q || 2 =2o-y pT y q 。The Hamming distance between the data point x p and x q matches the affinity AFF(p,q) defined above.

[0104] As Figure 3 shown in the block diagram of using multi-dimensional spectral hashing for the randomized palm vein feature vector according to an embodiment of the present invention, the steps of generating the first palm vein feature template are as follows:

[0105] S301. Calculate for each column in the randomized palm vein feature vector, and calculate the one-dimensional feature function φ ij (x(i)):

[0106]

[0107] where i represents the number of the first column in the randomized palm vein feature vector; j represents the number of the second column in the randomized palm vein feature vector; a i represents the minimum value in the column data of the randomized palm vein feature vector; b i represents the maximum value in the column data of the randomized palm vein feature vector; x(i) represents an arbitrary one-dimensional real feature corresponding to the i-th coordinate of the randomized palm vein feature vector x.

[0108] S302. Calculate the eigenvalue λ ij corresponding to each column data;

[0109]

[0110] where e represents the natural constant; δ 2 represents the variance of each column data.

[0111] S303. Select the column elements corresponding to several largest eigenvalues, rearrange the column elements in the order of the eigenvalue magnitudes to form a set A, and generate the first palm vein feature template.

[0112] Using the spectral graph partitioning technology of multi-dimensional spectral hashing, represent the palm vein data as a graph, where the palm vein feature data points are the nodes in the graph, and the affinity matrix represents the similarity between the nodes. The goal of spectral graph partitioning is to assign the similar palm vein feature vector nodes to the same set, so that the generated hash codes have similar representations. This ensures that the palm vein feature vectors of the same individual can generate similar feature templates, and the feature templates of the palm vein feature vectors of different individuals can maintain good performance.

[0113] S104. Perform a non-linear transformation on the first palm vein feature template using a non-linear activation function to obtain a second palm vein feature template.

[0114] In this embodiment, the method of non-linear transformation is as follows:

[0115] y2 = q(y1) = q(φ ij (B))

[0116]

[0117] where q(x) represents a non-linear activation function; tanh represents the hyperbolic tangent function; φ ij (B) represents a one-dimensional feature function; α represents a label information weight parameter, α = 0.4; y1 represents the first palm vein feature template; y2 represents the second palm vein feature template; B represents a feature matrix obtained by rearranging the column elements of the randomized palm vein feature vector X3 that satisfy the set A index.

[0118] Specifically, after using multi-dimensional spectral hashing, a palm vein template with good performance is obtained from the high-dimensional palm vein feature vector. Although multi-dimensional spectral hashing retains the accuracy of palm vein features, due to similarity preservation, it is vulnerable to similarity attacks. A similarity attack essentially exploits the similarity preservation property of data distribution to recover the original biometric. Conceptually, a similarity attack (Similarity Attack) attempts to optimize the reconstruction problem, i.e., arg min||u - u"||, where u represents the original biometric and u" represents the attacker's estimate of the original biometric u. Through the similarity preservation property, ||u - u"||≈||v - v"|| can be estimated from their hash values, where v represents the hashed template and v" represents the estimated hash template. If this estimation is accurate enough, a similar u" can be reconstructed. In short, similarity preservation may lead to information leakage, so it is very likely to retrieve the distance in the original biometric space from the hash space.

[0119] Therefore, after multi-dimensional spectral hashing, by performing a non-linear transformation on the palm vein feature template, the intra-class and inter-class distances of the template are affected. The palm vein feature template can meet the requirements of mutual information within a certain range, so that the feature template generated by the proposed method has the ability to resist similarity attacks.

[0120] After using the non-linear activation function, the palm vein data in the multi-dimensional space is mapped to a new coordinate system. Among them, the difference points that were originally far from the main cluster are transferred to the uniformly distributed main cluster by continuous periodic mapping. By this transfer, the large difference points are reduced, so that most data points are close together, thus reducing the inter-class distance and completing the additional protection against similarity attacks. Of course, when the similarity of the palm vein data distribution before and after is destroyed, the decrease in accuracy is inevitable. However, here according to the distribution characteristics of the palm vein data points, by adjusting the parameters of the function, better performance retention can be achieved.

[0121] S105. Convert the second palm vein feature template into a binary form to obtain a third palm vein feature template.

[0122] In this embodiment, after multi-dimensional spectral hashing and non-linear activation function are completed, a many-to-one sign function is used to convert the palm vein feature template into a final binary template. The specific method is as follows:

[0123] y3 = sign(y2) = sign(q(φ ij (B)))

[0124] where y3 represents the third palm vein feature template, and sign represents the binary sign function.

[0125] In addition to meeting the performance requirements of the template protection method and resisting similarity attacks, the transformed third palm vein feature template also has strong irreversibility.

[0126] According to the embodiments of the present invention, the initial palm vein features have strong concealment due to the many-to-one function and non-linear activation function. Compared with other palm vein template protection schemes, it can resist similarity attacks while maintaining performance loss, and has higher security. Due to the use of the spectral segmentation technique, the similarity structure of the palm vein data is preserved and concentrated. The accuracy of the palm vein features is well maintained, reducing the restriction of high security on performance, and achieving a balance between security and performance relative to the existing palm vein template protection schemes. Due to the user-specific random projection, the template is revocable. At the same time, bitwise operations can be performed for fast matching, and the binary hash code can also be used as a search index. In addition, non-linear general hashing is an unsupervised method based on statistical theory and is easy to implement.

[0127] It should be noted that for the foregoing method embodiments, for the sake of simple description, they are all expressed as a series of action combinations. However, those skilled in the art should know that the present invention is not limited by the described action sequence, because according to the present invention, certain steps can be performed in other sequences or simultaneously. Secondly, those skilled in the art should also know that the embodiments described in the specification are all optional embodiments, and the actions and modules involved are not necessarily essential to the present invention.

[0128] The above is the introduction of the method embodiments. The following is a further description of the solution of the present invention through an apparatus embodiment having the same inventive concept as the method in the foregoing embodiments.

[0129] As Figure 4 shown, the apparatus 400 includes:

[0130] S410. An acquisition module, configured to acquire original palm vein data and obtain an initial palm vein feature vector according to the original palm vein data.

[0131] S420, a random projection module, is used to perform random projection on the initial palm vein feature vector to obtain a randomized palm vein feature vector.

[0132] S430, a first template generation module, is used to generate a first palm vein feature template by using multi-dimensional spectral hashing on the randomized palm vein feature vector.

[0133] S440, a second template generation module, is used to perform a non-linear transformation on the first palm vein feature template by using a non-linear activation function to obtain a second palm vein feature template.

[0134] S450, a third template generation module, is used to perform binary conversion on the second palm vein feature template to obtain a third palm vein feature template.

[0135] Those skilled in the art can clearly understand that for the convenience and conciseness of description, the specific working processes of the described modules can refer to the corresponding processes in the foregoing method embodiments, and will not be elaborated herein.

[0136] In the technical solution of the present invention, the acquisition, storage, and application of the user's personal information involved all comply with the provisions of relevant laws and regulations and do not violate public order and good customs.

[0137] According to an embodiment of the present invention, the present invention also provides an electronic device and a readable storage medium.

[0138] Figure 5 The schematic block diagram of an electronic device 500 that can be used to implement the embodiments of the present invention is shown. The electronic device is intended to represent various forms of digital computers, such as, a laptop computer, a desktop computer, a workbench, a personal digital assistant, a server, a blade server, a mainframe computer, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as, a personal digital processor, a cellular phone, a smart phone, a wearable device, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are only examples and are not intended to limit the implementation of the present invention described and / or claimed herein.

[0139] The electronic device 500 includes a computing unit 501, which can execute various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 502 or a computer program loaded from a storage unit 508 into a random access memory (RAM) 503. In the RAM 503, various programs and data required for the operation of the electronic device 500 can also be stored. The computing unit 501, the ROM 502, and the RAM 503 are connected to each other through a bus 504. An input / output (I / O) interface 505 is also connected to the bus 504.

[0140] Multiple components in the electronic device 500 are connected to the I / O interface 505, including: an input unit 506, such as a keyboard, a mouse, etc.; an output unit 507, such as various types of displays, speakers, etc.; a storage unit 508, such as a disk, an optical disc, etc.; and a communication unit 509, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 509 allows the electronic device 500 to exchange information / data with other devices via a computer network such as the Internet and / or various telecommunication networks.

[0141] The computing unit 501 can be various general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 501 include but are not limited to a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 501 executes the various methods and processes described above, such as methods S101 - S105. For example, in some embodiments, methods S101 - S105 can be implemented as a computer software program, which is tangibly contained in a machine-readable medium, such as the storage unit 508. In some embodiments, part or all of the computer program can be loaded and / or installed onto the electronic device 500 via the ROM 402 and / or the communication unit 509. When the computer program is loaded into the RAM 503 and executed by the computing unit 501, one or more steps of the methods S101 - S105 described above can be executed. Alternatively, in other embodiments, the computing unit 501 can be configured to execute methods S101 - S105 in any other suitable manner (e.g., by means of firmware).

[0142] The various embodiments of the systems and technologies described above in this article can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGA), application-specific integrated circuits (ASIC), application-specific standard products (ASSP), system-on-a-chip (SOC), complex programmable logic devices (CPLD), computer hardware, firmware, software, and / or combinations thereof. These various embodiments can include: implemented in one or more computer programs, the one or more computer programs can be executed and / or interpreted on a programmable system including at least one programmable processor, the programmable processor can be a dedicated or general-purpose programmable processor, can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit the data and instructions to the storage system, the at least one input device, and the at least one output device.

[0143] The program code for implementing the method of the present invention can be written in any combination of one or more programming languages. These program codes can be provided to the processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing devices, such that when the program codes are executed by the processor or controller, the functions / operations specified in the flowchart and / or block diagram are implemented. The program codes can be executed entirely on the machine, partially on the machine, executed partially on the machine as an independent software package and partially on a remote machine, or executed entirely on a remote machine or server.

[0144] In the context of the present invention, a machine-readable medium can be a tangible medium that can contain or store a program for use by or in connection with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of a machine-readable storage medium would include an electrical connection based on one or more wires, a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0145] In order to provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and a pointing device (e.g., a mouse or a trackball) through which the user can provide input to the computer. Other kinds of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and the input received from the user can be in any form (including acoustic input, speech input, or tactile input).

[0146] The systems and techniques described herein can be implemented in a computing system including backend components (e.g., as a data server), or a computing system including middleware components (e.g., an application server), or a computing system including frontend components (e.g., a user computer having a graphical user interface or a web browser through which a user can interact with an implementation of the systems and techniques described herein), or a computing system including any combination of such backend components, middleware components, or frontend components. The components of the system can be interconnected to each other by digital data communication in any form or medium (e.g., a communication network). Examples of communication networks include: local area network (LAN), wide area network (WAN), and the Internet.

[0147] A computer system can include a client and a server. The client and the server are generally remote from each other and typically interact through a communication network. The client-server relationship is generated by computer programs running on the respective computers and having a client-server relationship with each other. The server can be a cloud server, a server of a distributed system, or a server incorporating a blockchain.

[0148] It should be understood that various forms of the processes shown above can be used, with steps reordered, added, or deleted. For example, the steps recited in the present invention can be executed in parallel, sequentially, or in a different order, as long as the desired results of the technical solution of the present invention can be achieved, and no limitation is imposed herein.

[0149] The above specific embodiments do not constitute a limitation on the protection scope of the present invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A palm vein feature template generation method for resisting similarity attacks, characterized in that: include: Acquire original palm vein data, and obtain an initial palm vein feature vector according to the original palm vein data; Performing random projection on the initial palm vein feature vector to obtain a randomized palm vein feature vector; Using multi-dimensional spectral hashing on the randomized palm vein feature vector to generate a first palm vein feature template; Using a nonlinear activation function to perform a nonlinear transformation on the first palm vein feature template to obtain a second palm vein feature template; Performing binary conversion on the second palm vein feature template to obtain a third palm vein feature template; The randomly projecting the initial palm vein feature vector comprises: Generate a pseudo-random matrix using an agreed key; The pseudo-random matrix is ​​converted into a standard orthogonal matrix using Schmidt orthogonalization method; Inputting the initial palm vein feature vector into the standard orthogonal matrix to obtain a randomized palm vein feature vector; The step of using multi-dimensional spectral hashing on the randomized palm vein feature vector to generate a first palm vein feature template includes: First, calculate each column in the randomized palm vein feature vector to calculate the one-dimensional feature function φ ij (x(i)): Where i represents the first column number in the randomized palm vein feature vector; j represents the second column number in the randomized palm vein feature vector; a i represents the minimum value in the column data of the randomized palm vein feature vector; b i represents the maximum value in the column data of the randomized palm vein feature vector; x(i) represents a one-dimensional arbitrary real feature corresponding to the i-th coordinate of the randomized palm vein feature vector x; Secondly, calculate the eigenvalue λ corresponding to each column data ij ; Among them, e represents a natural constant; δ 2 Indicates the variance of each column of data; Finally, select the largest number of eigenvalues ​​corresponding to the out-of-column elements, rearrange the column elements in the order of eigenvalue size to form a set A, and generate the first palm vein feature template; The step of performing a nonlinear transformation on the first palm vein feature template by using a nonlinear activation function to obtain a second palm vein feature template includes: y2=q(y1)=q(φ ij (B)) Among them, q(φ ij (B)) represents a nonlinear activation function; tanh is the hyperbolic tangent function; φ ij (B) represents a one-dimensional feature function; α represents a label information weight parameter, α=0.4; y1 represents the first palm vein feature template; y2 represents the second palm vein feature template; B represents a feature matrix in which the column elements satisfying the index of set A in the randomized palm vein feature vector X3 are rearranged.

2. The method according to claim 1, characterized in that The step of obtaining an initial palm vein feature vector according to the original palm vein data comprises: Mapping the original palm vein data to the trained projection matrix to obtain an initial palm vein feature vector; The trained projection matrix is ​​obtained by iterative calculation using the SDSPCA-NPE method.

3. The method according to claim 1, characterized in that The step of inputting the initial palm vein feature vector into the standard orthogonal matrix to obtain a randomized palm vein feature vector comprises: X3=X2×R1,X3∈ n Among them, X3 represents the randomized palm vein feature vector; X2 represents the initial palm vein feature vector; R1 represents the standard orthogonalized pseudo-random matrix; represents the real number set; n represents the dimension of the vector.

4. The method according to claim 1, characterized in that: The step of performing binary conversion on the second palm vein feature template to obtain a third palm vein feature template includes: y3=sign(y2)=sign(q(φ ij (B))) Wherein, y3 represents the third palm vein feature template, and sign represents the binary sign function.

5. A palm vein feature template generation device for resisting similarity attacks, characterized in that: include: An acquisition module, used for acquiring original palm vein data, and obtaining an initial palm vein feature vector according to the original palm vein data; A random projection module, used for randomly projecting the initial palm vein feature vector to obtain a randomized palm vein feature vector; A first template generating module, configured to generate a first palm vein feature template by using multi-dimensional spectrum hashing on the randomized palm vein feature vector; A second template generating module, configured to perform a nonlinear transformation on the first palm vein feature template using a nonlinear activation function to obtain a second palm vein feature template; The third template generating module is used to perform binary conversion on the second palm vein feature template to obtain a third palm vein feature template.

6. An electronic device comprising at least one processor; and A memory communicatively connected to the at least one processor; characterized in that The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method according to any one of claims 1 to 4.

7. A non-transitory computer-readable storage medium storing computer instructions, characterized in that: The computer instructions are used to cause the computer to execute the method according to any one of claims 1-4.

Images