Method and device for palmprint template protection and recognition based on multimodal shared key

Abstract

The present invention relates to the field of biometric recognition technology, and more specifically, to a palm template protection and recognition method and device based on a multi-modal shared key. The present invention generates a random key that is not related to the original biometric feature, and then encrypts it to form a biometric template, thereby realizing and strengthening the irreversibility of the feature; also constructs and trains a shared key generation network to transform palmprint and palm vein into the same feature space, while realizing end-to-end encryption of the biometric template, improving the correlation between modalities and the security of the template; also introduces LDPC encoding and decoding to realize key error correction, reduce noise interference, and improve the discriminability of the key. The present invention can simultaneously improve the discriminability, correlation, and security of multi-modal features, and the recognition rate is also improved, solving the problems that existing methods do not integrate template protection and the recognition rate is not ideal in some cases.

Description

Technical Field

[0001] The present invention relates to the field of biometric recognition technology, and more specifically, to a palm template protection and recognition method based on a multi-modal shared key, and a device using the method. Background Art

[0002] Biometric recognition has been widely used due to its convenience. User biometrics are permanent and unchangeable. Once leaked, it will cause irreversible huge losses to users. Therefore, research on protecting stored templates has emerged, namely biometric template protection (BTP), which includes cancelable biometrics (CB) and biometric cryptosystems (BCS).

[0003] Palm features are obvious, and it is easy to recover part or all of the original features through reversible attacks, resulting in permanent leakage of user features. Objectively, it is very necessary to protect the template.

[0004] Existing biometric recognition pays more attention to improving the recognition rate and uses additional methods to complete template protection, so that BCS is generally designed in a non-end-to-end manner. At the same time, palm veins and palm prints have a certain correlation. In existing methods, the feature data of each modality are still independent of each other and do not fully reflect the correlation between modalities, resulting in unsatisfactory recognition rates in multi-modal and cross-modal situations. Summary of the Invention

[0005] Based on this, it is necessary to provide a palm template protection and recognition method and device based on a multi-modal shared key to address the problems that existing methods do not integrate template protection and the recognition rate is not ideal in some cases.

[0006] The present invention is implemented by the following technical solutions:

[0007] In a first aspect, the present invention discloses a palm template protection and recognition method based on a multi-modal shared key, which is used to perform template protection and matching recognition on the palms of m target users according to a preset mode. The preset mode includes a multi-modal mode, a single-modal mode, and a cross-modal mode.

[0008] The palm template protection and recognition method based on a multi-modal shared key includes the following steps:

[0009] Step 1, use the trained key generation network to receive the i-th real-time palm print image or / and real-time palm vein image and generate a shared key

[0010] Among them, the historical palmprint images and historical palm vein images of m target users are used to train the key generation network, and the trained key generation network is obtained;

[0011] In the multi-modal mode or cross-modal mode, including the shared palmprint key and the shared palm vein key In the single-modal mode, it is the shared palmprint key or the shared palm vein key

[0012] Step 2, perform LDPC decoding and correction on to obtain the corrected key

[0013] Step 3, encrypt through SHA-256 to obtain the encrypted key and perform matching and recognition according to ;

[0014] In the multi-modal mode or single-modal mode, m random keys k are also pre-generated; among them, the j-th random key corresponds to the j-th target user, j ∈ [1, m]; then is encrypted through SHA-256 to obtain the biometric template Compare with If then the matching and recognition are successful; among them, in the multi-modal mode, including the encrypted palmprint key and the encrypted palm vein key In the single-modal mode, it is the encrypted palmprint key or the encrypted palm vein key;

[0015] In the cross-modal mode, including the encrypted palmprint key and the encrypted palm vein key Compare with If the Hamming distance between is less than the preset threshold, then the matching and recognition are successful.

[0016] The palm template protection and recognition method based on multi-modal shared keys implements the method or process according to the embodiments of the present disclosure.

[0017] In a second aspect, the present invention discloses a palm template protection and recognition device based on a multi-modal shared key, including: a random key generation module, a SHA-256 encryption module I, a shared key generation module, an LDPC decoding module, a SHA-256 encryption module II, and a matching and recognition module.

[0018] The random key generation module is used to pre-generate m random keys k in a multi-modal mode or a single-modal mode; where the j-th random key corresponds to the j-th target user, and j ∈ [1, m]. The SHA-256 encryption module I is used to obtain a biometric template after SHA-256 encryption The shared key generation module is used to receive the i-th real-time palmprint image and / or real-time palm vein image using a trained key generation network to generate a shared key The LDPC decoding module is used to perform LDPC decoding and correction to obtain a corrected key The SHA-256 encryption module II is used to obtain an encrypted key after SHA-256 encryption The matching and recognition module is used to perform matching and recognition based on

[0019] The palm template protection and recognition device based on the multi-modal shared key implements the method or process according to the embodiments of the present disclosure.

[0020] In a third aspect, the present invention discloses a readable storage medium. Computer program instructions are stored in the readable storage medium. When the computer program instructions are read and run by a processor, the steps of the method for protecting and recognizing a palm template based on a multi-modal shared key disclosed in the first aspect are executed.

[0021] Compared with the prior art, the present invention has the following beneficial effects:

[0022] 1. The present invention realizes the irreversibility of features by generating random keys that are not related to the original biometric features; by constructing and training a key generation network, the palmprint and palm vein are transformed into the same feature space, so that the generated shared key is bound to the biometric template of the target user, which can ensure end-to-end encryption of the biometric template while enhancing the correlation between modalities and the security of the template.

[0023] 2. The present invention uses SHA-256 for encryption to protect the key and strengthen the irreversibility of the features. And LDPC encoding and decoding are introduced to correct the key and reduce noise interference, improving the discriminability of the key.

[0024] 3. The present invention can be applied to various situations such as multi-modal, single-modal, and cross-modal, and the recognition rate has also been improved. Moreover, it can be deployed flexibly and can adapt to a variety of application scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 It is a brief flowchart of the palmprint template protection and recognition method based on multi-modal shared keys in multi-modal mode or single-modal mode in Embodiment 1 of the present invention;

[0026] Figure 2 It is a brief flowchart of the palmprint template protection and recognition method based on multi-modal shared keys in cross-modal mode in Embodiment 1 of the present invention;

[0027] Figure 3 is Figure 1 corresponding data flow diagram;

[0028] Figure 4 is Figure 1 data flow diagram during the training of the key generation network of;

[0029] Figure 5 is Figure 2 corresponding data flow diagram. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0030] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0031] It should be noted that when a component is referred to as being "installed on" another component, it can be directly on the other component or there may also be an intermediate component. When a component is considered to be "set on" another component, it can be directly set on the other component or there may be an intermediate component at the same time. When a component is considered to be "fixed to" another component, it can be directly fixed to the other component or there may be an intermediate component at the same time.

[0032] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those of ordinary skill in the technical field to which the present invention belongs. The terms used in the description of the present invention herein are only for the purpose of describing specific embodiments and are not intended to limit the present invention. The term "or / and" used herein includes any and all combinations of one or more of the related listed items.

[0033] Embodiment 1

[0034] Please refer to Figure 1 and 2 , which is a brief flowchart of the palmprint template protection and recognition method based on multi-modal shared keys provided in this Embodiment 1 in a preset mode. The preset modes include multi-modal mode, single-modal mode, and cross-modal mode.

[0035] Among them, Figure 1 is in multi-modal mode or single-modal mode, Figure 2 is in cross-modal mode.

[0036] Generally speaking, the palmprint template protection and recognition method based on multi-modal shared keys provided in this Embodiment 1 is used to perform template protection and matching recognition on the palms of m target users according to the preset mode. It should be noted that the value of m can be set according to the actual situation and theoretically has no upper limit.

[0037] (1) Refer to Figure 3 , in multi-modal mode or single-modal mode, the palmprint template protection and recognition method based on multi-modal shared keys includes the following steps:

[0038] Step 1, use the trained key generation network (which can be written as PalmKeyNet) to receive the i-th real-time palmprint image or / and real-time palm vein image, and generate a shared key

[0039] First of all, for the key generation network, it needs to be pre-constructed and trained:

[0040] The key generation network includes a palmprint key generation sub-network and a palm vein key generation sub-network. Both sub-networks use a deep CNN network, such as ResNet18, EffNetb5, MobileNetV2, etc. The specific network to be selected depends on the application requirements and the limitations of computing resources - ResNet18 is suitable for applications with larger computing resources and higher accuracy requirements, EffNetb5 is suitable for scenarios that balance performance and computing resources, while MobileNetV2 focuses on lightweight and computing resource-constrained devices.

[0041] Specifically, refer to Figure 4 , the training method of the key generation network includes:

[0042] Pre-generate m random keys k; among them, the j-th random key corresponds to the j-th target user, j ∈ [1, m].

[0043] Encode using LDPC to obtain the encoded key And it is used as the true label for the training of the key generation network. This is because, during the feature data conversion process, noise interference will inevitably occur and affect the recognition performance, while LDPC is an error correction technology that can reduce the noise impact during the data conversion process.

[0044] Input the historical palmprint images and historical palmar vein images of m target users into the key generation network; among them, the historical palmprint images and historical palmar vein images of the j-th target user are from the same hand of the j-th target user.

[0045] The palmprint key generation sub-network maps the historical palmprint image of the j-th target user to the key space A and generates a training key The palmar vein key generation sub-network maps the historical palmar vein image of the j-th target user to the key space A and generates a training key

[0046] After N rounds of training, the network converges, Consistent with A trained key generation network is obtained.

[0047] The purpose of the above training is to bind the palmprint images and palmar vein images of the same identity to the same key space. In this way, feature fusion is achieved, which can not only increase the intra-class correlation between the two, but also the key irrelevant to the features can improve the key discriminability.

[0048] Then, after completing the training of the key generation network, use the trained key generation network to process the i-th real-time palmprint image or / and real-time palmar vein image to generate a shared key Since the trained key generation network has achieved key binding, then, if the i-th real-time palmprint image or / and real-time palmar vein image comes from the j-th target user, then Is Bound.

[0049] Specifically, in the multi-modal mode, the trained key generation network simultaneously receives the real-time palmprint image and the real-time palmar vein image, and generates the i-th shared palmprint key The i-th shared palmar vein key In this way, Including And

[0050] In the single-modal mode, the trained key generation network receives the real-time palmprint image of the i-th target user and generates the i-th shared palmprint key Or the trained key generation network receives the real-time palmar vein image of the i-th target user and generates the i-th shared palmar vein key In this way, For or

[0051] Step 2: Perform LDPC decoding and correction on to obtain the i-th corrected key

[0052] Since the object of LDPC decoding is binary code, therefore, first perform binarization on and then perform LDPC decoding processing. During the processing, correction is achieved to obtain Among them, LDPC decoding is to add several redundant codes for error correction to the binarized After error correction is completed, these redundant codes are discarded, so that no additional storage space is occupied.

[0053] Specifically, in the multimodal mode, first perform binarization on and then perform LDPC decoding and correction to obtain the i-th corrected palmprint key For first perform binarization and then perform LDPC decoding and correction to obtain the i-th corrected palm vein key In this way, includes and

[0054] In the single-modal mode, if the result obtained in Step 1 is For first perform binarization and then perform LDPC decoding and correction to obtain the i-th corrected palmprint key If the result obtained in Step 1 is For first perform binarization and then perform LDPC decoding and correction to obtain the i-th corrected palm vein key In this way, includes or

[0055] Step 3: Encrypt using SHA-256 to obtain the i-th encrypted key and perform matching and recognition based on This step is to perform matching and recognition. Based on the basic principle, the Hamming distance D

[0056] (X, Y) between the objects X and Y to be compared is calculated. According to different preset modes, the specific methods of matching and recognition are different. H (X, Y).

[0057] In the multimodal mode or the single-modal mode, The object to be compared is a biometric template

[0058] And the biometric template is obtained in the following way:

[0059] S1. Pre-generate m random keys k; among them, the j-th random key corresponds to the j-th target user, where j ∈ [1, m]. The m random keys k are generated corresponding to m target users. In this way, the j-th random key corresponds to the j-th target user.

[0060] It should be noted that the m random keys k are all irrelevant to the original biometrics (i.e., palmprint and palm vein). And the m random keys are orthogonal to each other and have a sufficient mutual distance, so as to provide a large enough difference, thus representing different target users. Therefore, it is impossible to reverse-infer the original biometrics from the random key k, realizing the irreversibility of the feature.

[0061] It should be noted that the m random keys k here are the same as the m random keys k in the training method of the key generation network.

[0062] S2. Then After being encrypted by SHA-256, the j-th biometric template is obtained Among them, SHA-256 is a secure and reliable hashing algorithm, with the characteristics of uniqueness and irreversibility. Using SHA-256 to encrypt is also an irreversible process, which can provide additional security protection and strengthen the irreversibility of the feature.

[0063] In addition, the j-th biometric template can be stored in the database for unified management, which is also convenient for subsequent retrieval.

[0064] In this way, in the multi-modal mode, is compared with If and That is to say, and Then the matching recognition is successful.

[0065] In the single-modal mode, is compared with If the result of step two is then is compared with On this premise, if That is Then the matching recognition is successful. If the result of step two is will compare with On this premise, if i.e., then the matching recognition is successful.

[0066] It should be noted that if is stored in the database, then will be matched and recognized one by one with all the keys in the database. Once the Hamming distance is equal to 0, it means that the

[0067] (2) Refer to Figure 5 , in the cross-modal mode, the palmprint template protection and recognition method based on multi-modal shared keys is mainly different from the multi-modal mode in (1) in step three:

[0068] First of all, there is no need to obtain the biometric template

[0069] Secondly, including the encrypted palmprint key and the encrypted palmprint key will compare with If the Hamming distance between is less than the preset threshold, then the matching recognition is successful. Among them, the preset threshold is configured according to the actual situation.

[0070] Generally speaking, the above 3 preset modes can correspond to different application scenarios:

[0071] Since the single-modal mode only requires one input and is more convenient, it is suitable for scenarios with ordinary-level security requirements;

[0072] Since the multi-modal mode requires two inputs, it is suitable for scenarios with higher security requirements;

[0073] And since the cross-modal mode does not match the data in the database, it can be used to only confirm the identity: for example, input the palm vein image of an unknown identity and several palmprint images of known identities, then this mode can be used to match and recognize which known identity the palm vein image belongs to.

[0074] Embodiment 2

[0075] This Embodiment 2 discloses a palmprint template protection and recognition device based on multi-modal shared keys, which uses the palmprint template protection and recognition method based on multi-modal shared keys in Embodiment 1.

[0076] The palmprint template protection and recognition device based on a multi-modal shared key includes: a random key generation module, a SHA-256 encryption module I, a shared key generation module, an LDPC decoding module, a SHA-256 encryption module II, and a matching and recognition module.

[0077] The random key generation module is used to pre-generate m random keys k in the multi-modal mode or the single-modal mode; among them, the j-th random key corresponds to the j-th target user, where j ∈ [1, m].

[0078] The SHA-256 encryption module I is used to encrypt the biometric template obtained after SHA-256 encryption In addition, it can be configured not to work in the cross-modal mode.

[0079] The shared key generation module is used to receive the i-th real-time palmprint image or / and real-time palm vein image using the trained key generation network and generate a shared key The LDPC decoding module is used to perform LDPC decoding and correction to obtain a corrected key The SHA-256 encryption module II is used to encrypt the encrypted key obtained after SHA-256 encryption The matching and recognition module is used to perform matching and recognition based on For the specific difference recognition method, refer to Embodiment 1, which will not be elaborated here.

[0080] If the model training function is also considered to be integrated into this device, this device may further include an LDPC encoding module and a model training module.

[0081] The random key generation module also pre-generates m random keys k during the training of the key generation network; among them, the j-th random key corresponds to the j-th target user, where j ∈ [1, m]. The LDPC encoding module is used to, during the training of the key generation network, perform LDPC encoding on the encoded key obtained after LDPC encoding

[0082] The model training module is used to use it as the true label for the training of the key generation network and train the key generation network to obtain a trained key generation network. For the specific training method of the key generation network, refer to Embodiment 1, which will not be elaborated here.

[0083] Embodiment 3

[0084] Embodiment 3 discloses a readable storage medium storing computer program instructions. When the computer program instructions are read and run by a processor, the method for protecting and recognizing palm templates based on multi-modal shared keys in Embodiment 1 is executed.

[0085] When the method in Embodiment 1 is applied, it can be applied in the form of software. For example, it can be designed as an independently running program on a computer-readable storage medium. The computer-readable storage medium can be a USB flash drive, designed as a USB key, and designed as a program that starts the entire method through external triggering via the USB flash drive.

[0086] Embodiment 4

[0087] Embodiment 4 verifies the method in Embodiment 1:

[0088] The inventors selected 3 publicly available databases: 1) PolyU - Multispectral (PolyU - M); 2) CASIA - Multispectral Palmprint V1 (Casia - M); 3) Tongji Palmprint / Palm Vein (Tongji). Among them, if the original database does not provide palmprints, the palmseg tool is used for extraction.

[0089] In Embodiment 4, the key generation network respectively selects ResNet18, EffNetb5, and MobileNetV2 as the backbone network. That is to say, Embodiment 4 provides 3 kinds of PalmKeyNets.

[0090] (1) Apply the above 3 kinds of PalmKeyNets to the above 4 databases for performance comparison.

[0091] For any of the above databases, the training set and the test set are divided in a 1:1 ratio. The key generation network is trained according to the above training method using the training set. Among them, when training the key generation network, the input images are all normalized. The test set is used to test the method for protecting and recognizing palm templates based on multi-modal shared keys as described above.

[0092] Embodiment 4 uses the Identification Rate (IR) to evaluate the recognition performance and the False Acceptance Rate (FAR) to evaluate the authentication performance. Among them, in the cross-modal mode, only its false acceptance rate is reported.

[0093] The comparison results are as follows in Table 1:

[0094] Table 1 Performance Comparison Results

[0095]

[0096]

[0097] In Table 1, Cross far represents the FAR of the cross-modal mode; Print IR represents the IR of the unimodal mode (palmprint vs. palmprint); Print far represents the FAR of the unimodal mode (palmprint vs. palmprint); Vein IR represents the IR of the unimodal mode (palm vein vs. palm vein); Vein far represents the FAR of the unimodal mode (palm vein vs. palm vein); Fusion IR represents the IR of the multi-modal mode; Fusion far represents the FAR of the multi-modal mode.

[0098] As can be seen from Table 1, the above three key generation networks all have good performance on the above three databases.

[0099] In the unimodal mode, PalmKeyNet using EffNetb5 achieved excellent results in both unimodal and multi-modal mode matching on the PolyU-M database. The recognition rates of the two unimodal modes both reached 99.967%, and the false acceptance rate reached 0%. The recognition rate for Casia-M was slightly worse, but still not less than 95%. Therefore, it is recommended to use PalmKeyNet using EffNetb5 in practical applications.

[0100] Although the recognition rate of the multi-modal mode did not increase significantly compared with the unimodal mode, the false acceptance rate decreased significantly, which also indicates the effectiveness of the method in Example 1.

[0101] The false acceptance rates of the above three key generation networks in the cross-modal mode are all very low, which also shows the feasibility of the cross-modal mode.

[0102] (2) Apply the above three PalmKeyNets and some existing methods to the above four databases for performance comparison.

[0103] Among them, the existing methods used the Palmnet method, the DHN method, and the LDBC method. The equal error rate (EER) of the same database was selected for comparison. The lower the EER index, the higher the recognition rate.

[0104] The comparison results are as follows in Table 2:

[0105] Table 2 EER comparison results

[0106]

[0107]

[0108] In Table 2, Feature length represents the feature length; EER of prints represents the equal error rate of unimodal (palmprint); EER of veins represents the equal error rate of unimodal (palm vein); EER of multi modality represents the equal error rate of multi-modal; EER of cross modality represents the equal error rate of cross-modal.

[0109] As can be seen from Table 2, compared with other solutions, PalmKeyNet using EffNetb5 has significantly reduced most of the EER indicators, especially on the PolyU-M database. This shows that the method of Example 1 is superior to the existing methods.

[0110] In addition, PalmKeyNet using EffNetb5 also has better accuracy compared to the other two PalmKeyNets.

[0111] The technical features of the above-described embodiments can be combined arbitrarily. For the sake of brevity of description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as the scope described in this specification.

[0112] The above-described embodiments merely represent several implementation manners of the present invention, and the description thereof is relatively specific and detailed. However, it should not be construed as a limitation on the scope of the invention patent. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present invention, several modifications and improvements can still be made, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the invention patent should be subject to the appended claims.

Claims

1. A palmprint template protection and recognition method based on a multi-modal shared key, which is used to perform template protection and matching recognition on the palms of m target users according to a preset mode, and is characterized in that, The preset modes include a multi-modal mode, a single-modal mode, and a cross-modal mode; The palmprint template protection and recognition method based on a multi-modal shared key includes the following steps: Step 1: Use the trained key generation network to receive the i-th real-time palmprint image and / or real-time palm vein image, and generate a shared key Among them, the key generation network is trained using the historical palmprint images and historical palm vein images of m target users to obtain a trained key generation network; In the multimodal mode or cross-modal mode, including a shared palmprint key and a shared palm vein key In the unimodal mode, it is a shared palmprint key or a shared palm vein key Step 2, perform LDPC decoding and correction on to obtain the corrected key Step 3: Take the encrypted key obtained after SHA-256 encryption and perform matching and recognition based on it. In the multimodal mode or the unimodal mode, m random keys k are also pre-generated; among them, the j-th random key corresponds to the j-th target user, j ∈ [1, m]; then the biometric template obtained after being encrypted by SHA-256 will be compared with If then the matching and recognition are successful; among them, in the multimodal mode,[[]] includes the encrypted palmprint key and the encrypted palm vein key In the unimodal mode,[[]] is the encrypted palmprint key or the encrypted palm vein key; In the cross-modal mode, including an encrypted palmprint key and an encrypted palm vein key will be compared with If the Hamming distance between is less than the preset threshold, the matching recognition is successful.

2. The palmprint template protection and recognition method based on a multi-modal shared key according to claim 1, wherein The key generation network includes: a palmprint key generation sub-network and a palm vein key generation sub-network; The training method of the key generation network includes: Pre-generate m random keys k; among them, the j-th random key corresponds to the j-th target user, j ∈ [1, m]; After LDPC encoding, an encoded key is obtained and used as the true label for network training of key generation; Input the historical palmprint images and historical palmar vein images of m target users into the key generation network; among them, the historical palmprint images and historical palmar vein images of the j-th target user are from the same hand of the j-th target user; the palmprint key generation sub-network maps the historical palmprint images of the j-th target user to the key space A and generates training keys The palmar vein key generation sub-network maps the historical palmar vein images of the j-th target user to the key space A and generates training keys After N rounds of training, the network converges, which is consistent with, to obtain the trained key generation network.

3. The method for palm template protection and recognition based on multi-modal shared key according to claim 2, wherein Both the palmprint key generation sub-network and the palm vein key generation sub-network adopt a deep CNN network.

4. The method for palm template protection and recognition based on a multimodal shared key according to claim 1 or 2, characterized in that, The m random keys k are orthogonal to each other and have a sufficient mutual distance.

5. The method for palm template protection and recognition based on multi-modal shared keys according to claim 1, wherein In Step 3, in the multimodal mode or the unimodal mode, is stored in the database, and is matched and identified one by one with all the keys in the database.

6. The palmprint template protection and recognition method based on multimodal shared keys according to claim 1, wherein, In the multi-modal mode or the cross-modal mode, In Step 1, the trained key generation network receives the real-time palmprint image and the real-time palm vein image simultaneously to generate a shared palmprint key Shared palm vein key In Step 2, for perform binarization first, and then perform LDPC decoding correction to obtain a corrected palmprint key For perform binarization first, and then perform LDPC decoding correction to obtain a corrected palm vein key In Step 3, the is encrypted by SHA-256 to obtain an encrypted palmprint key The is encrypted by SHA-256 to obtain an encrypted palmprint key 7. The palmprint template protection and recognition method based on multimodal shared keys according to claim 1, wherein In the single-modal mode, In step one, the trained key generation network only receives real-time palmprint images and generates shared palmprint keys In step two, for First, perform binarization, and then perform LDPC decoding correction to obtain the corrected palmprint key In step three, the After being encrypted by SHA-256, the encrypted palmprint key is obtained Or in step one, the trained key generation network only receives real-time palm vein images and generates shared palm vein keys In step two, for First, perform binarization, and then perform LDPC decoding correction to obtain the corrected palm vein key In step three, the After being encrypted by SHA-256, the encrypted palm vein key is obtained 8. A palmprint template protection and recognition device based on a multi-modal shared key, characterized in that, It uses the palmprint template protection and recognition method based on a multi-modal shared key as described in any one of claims 1-5; A shared key generation module, which is used to receive the i-th real-time palmprint image and / or real-time palm vein image by using a trained key generation network, and generate a shared key LDPC decoding module, which is used to perform LDPC decoding correction to obtain a corrected key SHA-256 encryption module two, which is used to obtain an encrypted key after SHA-256 encryption A matching and recognition module, which is used to perform matching and recognition according to ; A random key generation module, which is used to pre-generate m random keys k in a multimodal mode or a unimodal mode; wherein, the j-th random key corresponds to the j-th target user, j ∈ [1, m]; And SHA-256 encryption module 1, which is used to obtain a biometric template by encrypting k j 1 after SHA-256 encryption 9. The palmprint template protection and recognition device based on a multi-modal shared key according to claim 8, characterized in that, When training the key generation network, the random key generation module also pre-generates m random keys k; among them, the j-th random key corresponds to the j-th target user, where j ∈ [1, m]; The palmprint template protection and recognition device based on a multi-modal shared key further includes: An LDPC encoding module, which is used to, when training a key generation network, obtain an encoded key after LDPC encoding And A model training module, which is used to serve as the true label for training the key generation network and train the key generation network to obtain a trained key generation network.

10. A readable storage medium, characterized in that, The readable storage medium stores computer program instructions, and when the computer program instructions are read and run by a processor, the steps of the palmprint template protection and recognition method based on a multi-modal shared key as described in any one of claims 1-7 are executed.

Images