Quantum resistance data encryption system and method based on biological characteristics

Abstract

The invention discloses a quantum resistance data encryption system and method based on biological characteristics, and the method comprises the steps: a data collection module obtains various biological characteristics and environment data of a user in real time, carries out the filtering of a preprocessing module, achieves the optimal fusion of multi-dimensional biological information through a multi-modal fusion module, and carries out the encryption of the multi-dimensional biological information; and the feature extraction module dynamically adjusts an extraction strategy to generate a biological feature template. And the key generation module generates a dynamic key in real time according to the template, and performs quantum resistance encryption fragmentation on the original data through the data encryption module. And the fragment recombination module completes data restoration after verifying the consistency of the biological characteristics and the secret key. According to the method, multi-source biological and environmental information is fully fused, highly personalized and dynamic data encryption and secure recombination are realized, the data confidentiality and the anti-attack ability under the background of quantum computing are effectively improved, and the method is suitable for a data protection scene with a high security requirement.

Description

Technical Field

[0001] The present invention relates to the technical field of data encryption, and in particular, to a quantum-resistant data encryption system and method based on biometrics. Background Art

[0002] With the rapid development of information technology, the transmission and storage of data in fields such as the Internet, the Internet of Things, and mobile communications are becoming increasingly frequent, and data security issues are becoming increasingly prominent. Traditional data encryption means mainly rely on symmetric encryption or asymmetric encryption algorithms, such as AES, RSA, etc. However, although these encryption algorithms still have strong security under the current computing power, with the continuous advancement of quantum computing technology, their cracking efficiency will be significantly improved, making the existing encryption algorithms face the risk of being broken.

[0003] To address the threat posed by quantum computing, researchers have proposed quantum-resistant cryptosystems based on lattice cryptography and have tried to apply them to actual data encryption scenarios. However, the current quantum-resistant encryption technologies still have problems such as complex key management, poor user experience, and lack of flexibility. In addition, a single-dimensional key generation mechanism (such as relying only on a single factor such as fingerprint, iris, or password) is vulnerable to attacks such as forgery, imitation, and static leakage, and cannot meet the dynamic authentication and encryption requirements in high-security scenarios.

[0004] In recent years, biometric recognition technology has developed rapidly, and its application in identity recognition and access control is becoming increasingly widespread, with advantages such as uniqueness, stability, and difficulty in imitation. However, most existing solutions only use a single biometric as the basis for identity authentication, and fail to conduct collaborative analysis by combining multi-modal biometrics and environmental factors, making it difficult to achieve higher-dimensional identity security protection.

[0005] Therefore, there is an urgent need for an encryption system that integrates multiple biometrics and environmental perception information, has a dynamic key generation mechanism, and is quantum-resistant, so as to improve the security and practicality of data encryption and meet the data security requirements in the future quantum era. Summary of the Invention

[0006] Aiming at the deficiencies of the existing technology, the purpose of the present invention is to provide a quantum-resistant data encryption system and method based on biometrics for improving the security and practicality of data encryption.

[0007] To achieve the above purpose, the present invention provides the following technical solutions: A quantum-resistant data encryption system based on biometrics, comprising: A data acquisition module for real-time collecting multiple biometric data of a user and multiple environmental data in the environment where the user is located; A preprocessing module, connected to the data acquisition module, is used to perform data preprocessing on each of the biometric data and the environmental data according to a preset preprocessing algorithm to obtain filtered biometric data and filtered environmental data after processing. The filtered biometric data includes iris images, fingerprint images, heart rate, and blood oxygen, and the filtered environmental data includes light intensity, temperature, humidity, and background noise; A multimodal fusion module, connected to the preprocessing module, is used to perform data fusion on each of the filtered biometric data, and find the optimal matching relationship between each of the filtered biometric data based on a preset optimization algorithm during the fusion process, and fuse to obtain multimodal fusion biometric data; A feature extraction module, connected to the multimodal fusion module, is used to perform feature extraction on the multimodal fusion biometric data according to a preset feature extraction algorithm, and dynamically adjust the parameters of the feature extractor in the feature extraction algorithm according to each of the filtered environmental data during the feature extraction process to obtain a biometric template; A key generation module, connected to the feature extraction module, is used to generate a biometric dynamic key in real time according to the biometric template; A data encryption module, connected to the key generation module, is used to dynamically divide the original data into multiple data slices according to the biometric dynamic key, and encrypt each of the data slices using a lattice-based quantum-resistant algorithm to obtain encrypted slices; A slice recombination module, connected to the data encryption module, is used to provide a quantum-resistant slice recombination strategy, so that when the user provides the correct biometric data and the dynamic key, each of the encrypted slices is restored to the original data.

[0008] Furthermore, the preprocessing module includes: An adaptive filtering unit, used to suppress random noise in each of the biometric data according to the adaptive median filtering algorithm to obtain first filtered data; A multi-resolution analysis unit, used to decompose the images in each of the biometric data according to the multi-resolution method, and then restore the detailed information in the images according to the threshold denoising technique to obtain second filtered data; A deep denoising unit, used to perform image denoising and enhancement processing on the images in each of the biometric data according to a pre-trained deep learning algorithm to obtain third filtered data. The first filtered data, the second filtered data, and the third filtered data constitute the filtered biometric data.

[0009] Furthermore, the multimodal fusion module includes: A feature alignment unit, which is used to extract iris vector features and fingerprint vector features of the iris image and the fingerprint image according to the ResNet algorithm, and extract heart rate time-series features and blood oxygen time-series features of the heart rate and the blood oxygen according to the LSTM network algorithm. Furthermore, according to the principal component analysis algorithm, the heart rate time-series features, the blood oxygen time-series features, the iris vector features and the fingerprint vector features are unified to the same dimension; A feature standardization unit, connected to the feature alignment unit, which is used to perform normalization processing on the heart rate time-series features, the blood oxygen time-series features, the iris vector features and the fingerprint vector features according to the Z-score normalization method to obtain a plurality of normalized features, and the normalized features include a heart rate normalized feature, a blood oxygen normalized feature, an iris normalized feature and a fingerprint normalized feature; A fusion unit, connected to the feature standardization unit, which is used to configure initial weights for each of the normalized features, introduce an attention mechanism to perceive the feature quality of each of the normalized features, and dynamically adjust the corresponding initial weights according to the feature quality to obtain adjusted weights. Finally, the multi-modal fusion biological data is calculated according to each of the normalized features and the corresponding adjusted weights.

[0010] Furthermore, the calculation formula of the multi-modal fusion biological data is configured as:

[0011] Wherein, represents the multi-modal fusion biological data, represents the normalized heart rate feature, represents the normalized blood oxygen feature, represents the normalized iris feature, represents the normalized fingerprint feature, represents the adjusted weight of the heart rate feature, represents the adjusted weight of the blood oxygen feature, represents the adjusted weight of the iris feature, represents the adjusted weight of the fingerprint feature, represents a preset offset constant, represents the Gamma function, represents the error function, represents the first kind of Bessel function, represents the natural logarithm function, represents the summation upper limit, represents the summation index variable, represents the summation denominator exponent parameter.

[0012] Furthermore, the multimodal fusion module further includes an optimization matching unit, connected to the fusion unit, for searching for the feature combination with the strongest correlation among the normalized features according to the maximum mutual information algorithm as the optimal matching relationship, and adjusting the calculation formula of the multimodal fusion biological data according to each of the normalized features and the corresponding adjustment weights in the optimal matching relationship, so as to obtain the optimized multimodal fusion biological data.

[0013] Furthermore, the feature extraction module includes: A feature extraction unit, configured to introduce a multi-layer residual convolution network, and perform feature extraction on the multimodal fusion biological data according to the multi-layer residual convolution network to obtain an intermediate feature tensor; A scoring and modeling unit, configured to construct an environmental scoring model based on a shallow neural network, and input the light intensity, the temperature, the humidity, and the background noise into the environmental scoring model to obtain a parameter adjustment factor; A parameter adjustment unit, connected to the scoring and modeling unit, for dynamically adjusting the convolution kernel coefficient of the feature extractor in the feature extraction algorithm according to the parameter adjustment factor to obtain an optimized extraction algorithm model, and the optimized extraction algorithm model is used to output an optimized feature tensor; A template generation unit, connected to the parameter adjustment unit and the feature extraction unit, for generating the biological feature template according to the optimized feature tensor.

[0014] Furthermore, the calculation formula of the parameter adjustment factor is configured as:

[0015] Wherein, represents the parameter adjustment factor, and respectively represent two weight matrices of the environmental scoring model, and respectively represent two bias terms of the environmental scoring model, respectively represent the light intensity, the temperature, the humidity, and the background noise, represents an activation function, represents the hyperbolic tangent function; The convolution kernel coefficient adjustment formula is configured as:

[0016] Wherein, represents the adjusted convolution kernel coefficient, represents the convolution kernel parameter before adjustment, represents the number of parameters to be adjusted, represents the each of the parameter adjustment factors.

[0017] Further, the filtered environmental data further includes a sensor angle, a physical occlusion amount, and a motion jitter amount, and the data encryption module includes: a perturbation generation unit configured to generate a perturbation vector based on the sensor angle, the physical occlusion amount, and the motion jitter amount, where the perturbation vector is used to control parameter settings in the data encryption process; a function generation unit connected to the perturbation generation unit and configured to generate a perturbation adjustment function according to the perturbation vector; a data encryption unit connected to the function generation unit and configured to dynamically adjust the number of shards, the number of encryption rounds, and the configuration of password parameters in the data encryption process according to the perturbation encryption function.

[0018] Further, the perturbation adjustment function is configured as:

[0019] wherein, represents the perturbation adjustment function, represents the perturbation vector, and represent two adjustable initial constants, respectively represent the sensor angle, the physical occlusion amount, and the motion jitter amount after normalization, represent three preset weight coefficients.

[0020] A quantum-resistant data encryption method based on biometrics is applied to the above-mentioned quantum-resistant data encryption system, and includes: Step S1, a data acquisition module collects multiple biometric data of a user and multiple environmental data in the environment where the user is located in real time; Step S2, a preprocessing module performs data preprocessing on each of the biometric data and the environmental data according to a preset preprocessing algorithm to obtain processed filtered biometric data and filtered environmental data, where the filtered biometric data includes an iris image, a fingerprint image, a heart rate, and a blood oxygen level, and the filtered environmental data includes an illumination intensity, a temperature, a humidity, and background noise; Step S3, a multimodal fusion module performs data fusion according to each of the filtered biometric data, and searches for an optimal matching relationship between each of the filtered biometric data based on a preset optimization algorithm during the fusion process, and fuses to obtain multimodal fusion biometric data; Step S4, a feature extraction module extracts features from the multimodal fusion biometric data according to a preset feature extraction algorithm, and dynamically adjusts parameters of a feature extractor in the feature extraction algorithm according to each of the filtered environmental data during the feature extraction process to obtain a biometric template; Step S5, the key generation module generates a biometric dynamic key in real time according to the biometric template; the data encryption module dynamically divides the original data into multiple data slices according to the biometric dynamic key, and encrypts each data slice by using a lattice-based quantum-resistant algorithm to obtain encrypted slices; Step S6, the slice recombination module provides a quantum-resistant slice recombination strategy, so that when the user provides the correct biometric data and the dynamic key, each encrypted slice is restored to the original data.

[0021] Advantages of the present invention: The present invention collects multiple biometric data such as the user's iris, fingerprint, heart rate, and blood oxygen, and combines external environmental data such as ambient light, temperature, humidity, and noise to dynamically generate a biometric dynamic key with strong uniqueness and high timeliness. This mechanism effectively prevents the key from being forged or replayed, and improves the overall security and reliability of the data encryption system; At the same time, the present invention uses a lattice-based encryption algorithm to encrypt data slices, which has natural resistance to quantum computing attacks, solves the problem that traditional encryption algorithms are vulnerable to being cracked in a quantum computing environment, and significantly enhances the forward-looking and adaptability of the system; The present invention integrates biometric information from multiple different dimensions, and performs matching and integration through an optimized algorithm, which not only improves the accuracy of feature recognition, but also enhances the robustness of the system to abnormal data or attack behaviors, and avoids system paralysis caused by the failure of a single feature; In addition, the present invention dynamically adjusts the parameters of the extractor in the feature extraction algorithm, so that it can adapt to different environmental conditions such as light, humidity, and temperature, effectively reducing the influence of external interference on the recognition result, and improving the adaptability and usability of the system in practical applications; The present invention divides the original data into multiple data slices and encrypts them separately. Only when the correct biometric and key are provided can the restoration be completed through the slice recombination module, enhancing the anti-tampering and anti-leakage capabilities of the data during transmission and storage, and improving the security of data transmission. Description of the drawings

[0022] Figure 1 is a schematic structural diagram of the quantum-resistant data encryption system in the present invention; Figure 2 is a schematic structural diagram of the preprocessing module in the present invention; Figure 3 is a schematic structural diagram of the multi-modal fusion module in the present invention; Figure 4 is a schematic structural diagram of the feature extraction module in the present invention; Figure 5It is a schematic structural diagram of the data encryption module in the present invention; Figure 6 It is a schematic diagram of the steps of the quantum-resistant data encryption method in the present invention.

[0023] Reference numerals: 1, data acquisition module; 2, preprocessing module; 21, adaptive filtering unit; 22, multi-resolution analysis unit; 23, deep denoising unit; 3, multi-modal fusion module; 31, feature alignment unit; 32, feature standard unit; 33, fusion unit; 34, optimization matching unit; 4, feature extraction module; 41, feature extraction unit; 42, scoring modeling unit; 43, parameter adjustment unit; 44, template generation unit; 5, key generation module; 6, data encryption module; 61, perturbation generation unit; 62, function generation unit; 63, data encryption unit; 7, sharding and recombination module. Detailed implementation manners

[0024] The present invention will be further described in detail below with reference to the drawings and embodiments. The same components are denoted by the same reference numerals. It should be noted that the terms "front", "rear", "left", "right", "upper" and "lower" used in the following description refer to the directions in the drawings, and the terms "bottom surface" and "top surface", "inner" and "outer" refer to the directions towards or away from the geometric center of a specific component respectively.

[0025] Example 1, referring to Figure 1 , which is the first embodiment of the present invention. This embodiment provides a quantum-resistant data encryption system based on biometrics, which can improve the security and practicability of data encryption. It includes: A data acquisition module 1, configured to collect multiple biometric data of a user in real time and multiple environmental data in the environment where the user is located; A preprocessing module 2, connected to the data acquisition module 1, configured to perform data preprocessing on each of the biometric data and the environmental data according to a preset preprocessing algorithm to obtain filtered biometric data and filtered environmental data after processing. The filtered biometric data includes iris images, fingerprint images, heart rate, and blood oxygen, and the filtered environmental data includes light intensity, temperature, humidity, and background noise; A multi-modal fusion module 3, connected to the preprocessing module 2, configured to perform data fusion according to each of the filtered biometric data, and find the optimal matching relationship between the filtered biometric data based on a preset optimization algorithm during the fusion process, and fuse to obtain multi-modal fusion biometric data; The feature extraction module 4, connected to the multimodal fusion module 3, is configured to extract features from the multimodal fusion biological data according to a preset feature extraction algorithm, and dynamically adjust the parameters of the feature extractor in the feature extraction algorithm according to each piece of filtered environmental data during the feature extraction process, so as to obtain a biological feature template; The key generation module 5, connected to the feature extraction module 4, is configured to generate a biological dynamic key in real time according to the biological feature template; The data encryption module 6, connected to the key generation module 5, is configured to dynamically divide the original data into multiple data slices according to the biological dynamic key, and perform data encryption on each data slice by using a lattice-based quantum-resistant algorithm to obtain encrypted slices; The slice recombination module 7, connected to the data encryption module 6, is configured to provide a quantum-resistant slice recombination strategy, so that when the user provides the correct biological feature data and the dynamic key, each encrypted slice is restored to the original data.

[0026] Specifically, in this embodiment, the data acquisition module 1 integrates the following sensors: Iris recognition sensor: Adopt a near-infrared iris scanning module (such as IRIS ID iCAM7000) to collect high-precision iris images of users.

[0027] Fingerprint sensor: Adopt a capacitive fingerprint collector (such as FPC 1020) to obtain fingerprint images in real time.

[0028] Heart rate and blood oxygen sensor: Adopt a MAX30102 optoelectronic blood volume pulse wave sensor to continuously measure the user's heart rate and blood oxygen saturation.

[0029] Ambient light sensor: Adopt a BH1750 digital light sensor to detect the current light intensity (lux value).

[0030] Temperature and humidity sensor: Use an SHT31 digital temperature and humidity sensor to collect environmental temperature and humidity.

[0031] Noise sensor: Use a MEMS microphone (such as INMP441) to collect the background noise intensity (dB level) in the current environment.

[0032] Working principle of Embodiment 1: Data acquisition and preprocessing: Each sensor collects the user's physiological characteristics and the environmental data in real time, and performs cleaning and enhancement on the original data through preset algorithms such as image enhancement (CLAHE), filtering (adaptive median filtering), and denoising (deep learning algorithm), and outputs high-quality filtered biological data and filtered environmental data.

[0033] Multimodal Fusion and Feature Extraction: Iris texture, fingerprint ridge direction, pulse spectrum features, etc. are extracted through a fusion algorithm to form a unified biometric vector. At the same time, data such as ambient light, temperature, humidity, and noise are introduced as weight correction parameters to dynamically adjust the parameters of the feature extractor, making the extraction process environmentally robust.

[0034] Key Generation and Quantum-Resistant Encryption: Based on the biometric vector, hash compression is performed (such as using SHA3), and a dynamic key is generated in combination with a timestamp or a random salt value. Subsequently, the original data is encrypted in fragments under the guidance of this key, and each fragment is encrypted through a lattice cryptography algorithm (such as Kyber based on LWE), ensuring that even if some data is intercepted, the complete information cannot be restored.

[0035] Fragment Reassembly and Data Recovery: The user needs to provide real-time biometric information again, and the system determines whether the same dynamic key can be generated. If the verification passes, the data fragments can be decrypted and reassembled to restore the original data.

[0036] In this embodiment, keys are generated based on biometric information and environmental data in multiple dimensions, which has strong individual uniqueness and time dynamics, significantly enhancing the anti-counterfeiting and anti-replay attack capabilities of the system.

[0037] At the same time, a lattice-cryptography-based encryption algorithm is used for data encryption, which has quantum resistance, ensuring that even if a general-purpose quantum computer appears in the future, the key cannot be effectively cracked, improving the security of data encryption.

[0038] In addition, in this embodiment, multiple biometric information is cross-checked, reducing the recognition failure caused by single-sensor failure or data anomaly, and improving the stability and reliability of the system in complex environments. Finally, the dynamic introduction of environmental data makes the feature extraction process have adaptive capabilities, effectively reducing the recognition error caused by factors such as light, humidity, and noise.

[0039] Preferably, referring to Figure 2 , the preprocessing module 2 includes: An adaptive filtering unit 21 for suppressing random noise in each biometric data according to the adaptive median filtering algorithm to obtain first filtered data; A multi-resolution analysis unit 22 for decomposing the images in each biometric data according to the multi-resolution method, and then restoring the detailed information in the images according to the threshold denoising technique to obtain second filtered data; A deep denoising unit 23 for performing image denoising and enhancement processing on the images in each biometric data according to a pre-trained deep learning algorithm to obtain third filtered data. The first filtered data, the second filtered data, and the third filtered data constitute the filtered biometric data.

[0040] Specifically, in this embodiment, the biometric data includes iris images and fingerprint images. The resolution of the iris image is 640x480, and the format is grayscale. The resolution of the fingerprint image is 512x512, and the format is binary. Heart rate and blood oxygen are waveform / time series signals.

[0041] The adaptive filtering unit 21 performs filtering using an improved adaptive median filtering algorithm (AMF): First, analyze the local pixel variance of the image, automatically adjust the filter window size, and then suppress Gaussian noise and salt-and-pepper noise. The output of this algorithm is the first filtered data (image noise is reduced, edges are blurred). This algorithm can improve the basic clarity of the image and suppress burst noise. After this processing, the filter window size of the iris image is dynamically adjusted from 3x3 to 5x5, and the noise level drops by about 42%.

[0042] The multi-resolution analysis unit 22 uses the discrete wavelet transform algorithm (DWT) + threshold denoising algorithm (hard threshold): First, perform 3-layer wavelet decomposition on the filtered image, then use threshold compression to reduce noise in the detail coefficients, and finally reconstruct the image to obtain the second filtered data (image details are restored, high-frequency noise is removed). After this processing, the blurred details in the fingerprint image, such as ridge intersections, are clearer, and the signal-to-noise ratio of the image is increased by about 18%. This processing can restore the edge details of the image and improve the structural fidelity.

[0043] The deep denoising unit 23 selects a model based on the DnCNN (Denoising Convolutional Neural Network) convolutional neural network. The training method of this model includes: using 3000 synthetic noise training images for supervised training; adopting the MSE loss function and the Adam optimizer. The deep denoising unit 23 inputs the second filtered image into the network and outputs the third filtered data (clear structure, texture retained). Testing on the iris image shows that after processing with DnCNN, the SSIM value of the image is increased from 0.68 to 0.91. The final filtered biometric data is a fusion representation of the three image preprocessing results.

[0044] Preferably, referring to Figure 3 , the multi-modal fusion module 3 includes: The feature alignment unit 31 is used to extract the iris vector features and fingerprint vector features of the iris image and fingerprint image according to the ResNet algorithm, and extract the heart rate time series features and blood oxygen time series features of the heart rate and blood oxygen according to the LSTM network algorithm. Then, according to the principal component analysis algorithm, the heart rate time series features, blood oxygen time series features, iris vector features, and fingerprint vector features are unified to the same dimension; Feature standard unit 32, connected to the feature alignment unit 31, is used to normalize the heart rate time-series features, blood oxygen time-series features, iris vector features, and fingerprint vector features according to the Z-score normalization method to obtain multiple normalized features, including heart rate normalized features, blood oxygen normalized features, iris normalized features, and fingerprint normalized features; Fusion unit 33, connected to the feature standard unit 32, is used to configure initial weights for each normalized feature respectively, introduce an attention mechanism to perceive the feature quality of each normalized feature, and dynamically adjust the corresponding initial weights according to each feature quality to obtain adjusted weights. Finally, multi-modal fusion biometric data is calculated based on each normalized feature and the corresponding adjusted weights.

[0045] Specifically, in this embodiment, the iris image is a grayscale image with a resolution of 640x480, the fingerprint image is a binary image with a resolution of 512x512, the heart rate signal is a photoplethysmogram within 10 seconds, the sampling rate is 250Hz, and the blood oxygen signal is collected synchronously with the heart rate, unit: %.

[0046] The feature alignment unit 31 converts features of different modalities (image / timeseries) into vectors in the same dimensional space for easy fusion.

[0047] Image modality processing (iris + fingerprint): Use pre-trained ResNet-34 to extract image features; the numerical ranges of the output iris vector features and fingerprint vector features are both.

[0048] Timeseries modality processing (heart rate + blood oxygen): Use a two-layer LSTM (128 hidden units) to extract timeseries dynamic features; each signal outputs a 128-dimensional timeseries vector feature; the numerical ranges of the output heart rate time-series features and blood oxygen time-series features are both.

[0049] Feature alignment: Use principal component analysis (PCA) to uniformly compress / expand all features to the same dimension and output a feature vector of the same dimension.

[0050] The feature standard unit 32 is used to unify the numerical scales of different feature dimensions, avoid the dominant deviation of a certain modality feature on the fusion result, and ensure the unity of different feature numerical scales. The central processing unit normalizes each feature vector using the Z-score normalization method and outputs 4 normalized features: heart rate normalized features, blood oxygen normalized features, iris normalized features, and fingerprint normalized features.

[0051] The fusion unit 33 combines the feature quality of each modality, realizes dynamic weighted fusion through the attention mechanism, dynamically perceives the quality of each modality through the attention mechanism, enhances the robustness of the fusion result, and then improves the biometric key stability, individual uniqueness, and anti-tampering ability through multi-modal fusion.

[0052] Preferably, the calculation formula of the multimodal fusion biological data is configured as:

[0053] Wherein, represents the multimodal fusion biological data, represents the normalized heart rate feature, represents the normalized blood oxygen feature, represents the normalized iris feature, represents the normalized fingerprint feature, represents the adjustment weight of the heart rate feature, represents the adjustment weight of the blood oxygen feature, represents the adjustment weight of the iris feature, represents the adjustment weight of the fingerprint feature, represents the preset offset constant, represents the Gamma function, represents the error function, represents the Bessel function of the first kind, represents the natural logarithm function, represents the upper limit of summation, represents the summation index variable, represents the summation denominator exponential parameter.

[0054] Specifically, in this embodiment, the central processing unit is embedded with a specially designed software program, and this program performs operations according to the calculation formula of the multimodal fusion biological data, where the parameters , , , are all positive real numbers, which are used to adjust the weights of each biological feature in the fusion process. is the offset constant used to translate to ensure that the input value of the Gamma function is greater than zero. This program uses the math library to implement the Gamma function , the Bessel function of the first kind , the error function and the natural logarithm function for operations, and calculates the fused biological data . This value is a positive real number, and the larger the value, the higher the overall matching or fusion degree of the four normalized biological features; at the same time, in this embodiment, the central processing unit can be used in combination with a quantum-resistant data encryption module to perform subsequent quantum-resistant encryption processing on the fused data , and store the operation result in a secure database or transmit it to an authorized terminal, so as to realize applications such as identity authentication, access control, and data encryption protection; in the device, each group of sensors is equipped with an adaptive calibration module, which is used to compensate in real time for the interference of environmental factors such as external light, temperature, noise, and movement on the collected data, and ensure the normalized feature values , , , Accurate and stable; all modules of the entire system are interconnected through a high-speed data bus, and self-check and synchronous calibration are performed when the system starts, ensuring that data acquisition, normalization processing, mathematical operations, and subsequent encryption and data transmission can be accurately completed within a short time in practical applications. Through the above embodiments, the present invention realizes the fusion of normalized features of heart rate, blood oxygen, iris, and fingerprint through a complex multi-function mathematical model, and the output multi-modal fusion data can be used for precise identity recognition and quantum-resistant data encryption applications, and the numerical domain belongs to positive real numbers, and theoretically ∈(0, +∞), where a larger value represents a high degree of fusion and high consistency, suitable for high-security biological authentication and encryption requirements.

[0055] Preferably, the multi-modal fusion module 3 further includes an optimization matching unit 34 and a connection fusion unit 33, which are used to search for the feature combination with the strongest correlation between normalized features as the optimal matching relationship according to the maximum mutual information algorithm, and adjust the calculation formula of the multi-modal fusion biological data according to the normalized features and corresponding adjustment weights in the optimal matching relationship to obtain optimized multi-modal fusion biological data.

[0056] Specifically, in this embodiment, the optimization matching unit 34 uses the maximum mutual information algorithm to calculate the information correlation between any two normalized features, constructs a mutual information matrix, and then selects the feature combination with the maximum mutual information among all feature pairs, sets it as the optimal matching relationship, and then according to the specific normalized feature values , , , and the corresponding adjustment weights , , , dynamically adjust the above calculation formula of the multi-modal fusion biological data to optimize the multi-modal fusion biological data. By analyzing mutual information to identify the most coupled feature combination, more accurate data fusion can be achieved, the discriminability of the fused biological features can be enhanced, and the uniqueness and stability of user individual recognition can be effectively improved. Compared with the traditional fixed fusion strategy, this solution can dynamically select the fusion path according to the information intensity between the actually collected features, and has stronger adaptability and generalization ability to adapt to feature changes in different physiological states or acquisition environments.

[0057] Since the fusion result is more in line with the actual coupling structure between individual biometric features, the generated biometric dynamic key has a higher entropy value and unpredictability, effectively enhancing the security of the quantum-resistant cryptographic system.

[0058] The information-theoretic method (maximum mutual information) and lightweight deep fusion mechanism adopted in this embodiment do not rely on complex network structures, ensuring efficient operation on edge devices or embedded platforms and having good engineering feasibility.

[0059] Preferably, referring to Figure 4 , the feature extraction module 4 includes: A feature extraction unit 41, which is used to introduce a multi-layer residual convolutional network and extract features from the multi-modal fusion biometric data according to the multi-layer residual convolutional network to obtain an intermediate feature tensor; A scoring modeling unit 42, which is used to construct an environmental scoring model based on a shallow neural network, and input the light intensity, temperature, humidity, and background noise into the environmental scoring model to obtain a parameter adjustment factor; A parameter adjustment unit 43, connected to the scoring modeling unit 42, which is used to dynamically adjust the convolution kernel coefficient of the feature extractor in the feature extraction algorithm according to the parameter adjustment factor to obtain an optimized extraction algorithm model, and the optimized extraction algorithm model is used to output an optimized feature tensor; A template generation unit 44, connected to the parameter adjustment unit 43 and the feature extraction unit 41, which is used to generate a biometric template according to the optimized feature tensor.

[0060] Specifically, in this embodiment, the feature extraction unit 41 introduces a multi-layer residual convolutional neural network, and its basic structure is: composed of several stacked residual modules, and each residual module contains two 3×3 convolutional layers, a batch normalization (BatchNormalization) layer, and a ReLU activation function. The input multi-modal fusion biometric data (including the fusion results of heart rate, blood oxygen, iris image, and fingerprint image) first undergoes input layer normalization processing, and then sequentially passes through multiple residual modules to extract spatial and temporal related features, and finally outputs an intermediate feature tensor.

[0061] The scoring modeling unit 42 constructs a shallow neural network model as the environmental scoring model, whose input is the light intensity, temperature, humidity, and background noise collected in real time, and the structure includes an input layer, a hidden layer (8 neurons, with the activation function being ReLU), and an output layer (1 neuron, with the activation function being Sigmoid).

[0062] This scoring model outputs a parameter adjustment factor ∈ [0,1], and its calculation formula is as follows:

[0063] Among them, represents a parameter adjustment factor, and respectively represent two weight matrices of the environmental scoring model, and respectively represent two bias terms of the environmental scoring model, respectively represent light intensity, temperature, humidity, and background noise, represents an activation function, represents the hyperbolic tangent function.

[0064] The parameter adjustment unit 43 dynamically adjusts the convolution kernel parameters of the convolution layer in the feature extraction network using the parameter adjustment factor output by the scoring modeling unit 42. The specific method is as follows: For each convolution kernel parameter in the residual module, the following update is performed: The convolution kernel coefficient adjustment formula is configured as:

[0065] Among them, represents the adjusted convolution kernel coefficient, represents the convolution kernel parameter before adjustment, represents the number of parameters to be adjusted, represents the th parameter adjustment factor.

[0066] That is, in the model inference stage, the network is guided by the current environmental state to adaptively strengthen the feature importance, thereby improving the system robustness. The adjusted network is called an optimized feature extraction algorithm model, and an optimized feature tensor after adjustment is output.

[0067] The template generation unit 44 sends the optimized feature tensor into a compression - quantization - mapping process, and finally generates a biometric feature template that can be used for subsequent encryption processes. The structure of this template includes: the main feature hash encoding in the image domain, the sparse representation encoding of the physiological feature sequence, and the fusion identification meta - information.

[0068] In this embodiment, by introducing an environmental perception mechanism and a dynamic network adjustment strategy, it is possible to still extract a stable, accurate, and significantly individual - different biometric template in a real - use environment with light fluctuations, temperature and humidity changes, or noise interference, thereby greatly improving the security and reliability of the subsequent key generation and data encryption processes.

[0069] Example 2, refer to Figure 5, which is the second embodiment of the present invention. Different from the previous embodiment, this embodiment provides a perturbation generation unit 61, a function generation unit 62, and a data encryption unit 63, which can enhance the dynamic adaptability of the data encryption process and significantly improve the security strength and environmental robustness of the data encryption system. Among them, the filtered environmental data also includes the sensor angle, the physical occlusion amount, and the motion jitter amount. Then, the data encryption module 6 includes: The perturbation generation unit 61 generates a perturbation vector based on the sensor angle, the physical occlusion amount, and the motion jitter amount. The perturbation vector is used to control the parameter setting of the data encryption process; The function generation unit 62 is connected to the perturbation generation unit 61 and is used to generate a perturbation adjustment function according to the perturbation vector; The data encryption unit 63 is connected to the function generation unit 62 and is used to dynamically adjust the number of shards, the number of encryption rounds, and the password parameter configuration in the data encryption process according to the perturbation encryption function.

[0070] Preferably, the perturbation adjustment function is configured as:

[0071] Among them, represents the perturbation adjustment function, represents the perturbation vector, and represent two adjustable initial constants, respectively represent the sensor angle, the physical occlusion amount, and the motion jitter amount after normalization, represents three preset weight coefficients.

[0072] Working principle of Embodiment 2: The perturbation generation unit 61 is deployed on the three-axis gyroscope of the terminal device to obtain the sensor angle in real time, cooperate with the infrared or laser ranging module to obtain the physical occlusion amount between the target and the sensor, and calculate the root mean square of the motion jitter of the device in unit time in combination with the accelerometer data. The three indicators are respectively normalized to construct the perturbation vector . The function generation unit generates the perturbation adjustment function according to the perturbation vector . The data encryption unit executes the following dynamic strategy: Setting of the number of shards: , where represents the number of shards, represents the size of the original data, represents the benchmark shard size.

[0073] Adjustment of the number of encryption rounds: , where represents the basic number of encryption rounds, Indicates the adjusted encryption rounds; Dynamic expansion of lattice cipher parameters (lattice dimension): Wherein, represents the basic lattice dimension, and the value of the basic lattice dimension is set to 512, represents the lattice dimension after dynamic expansion.

[0074] The encryption operation is performed based on lattice ciphers (such as NTRU or LWE variants) to ensure resistance to quantum attacks. This embodiment introduces a dynamic perturbation function for environmental perturbation factors, which can effectively enhance the dynamic adaptability of the data encryption process. Especially in mobile devices, wearable devices or complex environments, it significantly improves the security strength and environmental robustness of the data encryption system.

[0075] A biometric-based quantum-resistant data encryption method, applied to the above-mentioned quantum-resistant data encryption system, referring to Figure 6 , includes: Step S1, the data acquisition module 1 continuously acquires multiple biometric data of the user and multiple environmental data in the environment where the user is located; Step S2, the preprocessing module 2 respectively performs data preprocessing on each biometric data and environmental data according to a preset preprocessing algorithm to obtain the filtered biometric data and filtered environmental data after processing. The filtered biometric data includes iris images, fingerprint images, heart rate, and blood oxygen. The filtered environmental data includes light intensity, temperature, humidity, and background noise; Step S3, the multimodal fusion module 3 performs data fusion based on each filtered biometric data. During the fusion process, it searches for the optimal matching relationship between each filtered biometric data based on a preset optimization algorithm, and fuses to obtain multimodal fusion biometric data; Step S4, the feature extraction module 4 performs feature extraction on the multimodal fusion biometric data according to a preset feature extraction algorithm, and dynamically adjusts the parameters of the feature extractor in the feature extraction algorithm according to each filtered environmental data during the feature extraction process to obtain a biometric template; Step S5, the key generation module 5 continuously generates a biometric dynamic key according to the biometric template; the data encryption module 6 dynamically divides the original data into multiple data slices according to the biometric dynamic key, and performs data encryption on each data slice using a quantum-resistant algorithm based on lattice ciphers to obtain encrypted slices; Step S6, the slice recombination module 7 provides a quantum-resistant slice recombination strategy to enable the user to restore each encrypted slice to the original data when providing the correct biometric data and dynamic key.

[0076] The above are only the preferred embodiments of the present invention, and the protection scope of the present invention is not limited to the above embodiments. All technical solutions falling within the concept of the present invention belong to the protection scope of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the principle of the present invention, several improvements and refinements should also be regarded as within the protection scope of the present invention.

Claims

1. A quantum-resistant data encryption system based on biometrics, characterized in that, Including: A data acquisition module (1) for real-time acquisition of multiple biometric data of a user and multiple environmental data in the environment where the user is located; A preprocessing module (2) connected to the data acquisition module (1) for performing data preprocessing on each of the biometric data and the environmental data according to a preset preprocessing algorithm to obtain filtered biometric data and filtered environmental data after processing. The filtered biometric data includes iris images, fingerprint images, heart rate, and blood oxygen, and the filtered environmental data includes light intensity, temperature, humidity, and background noise; A multimodal fusion module (3) connected to the preprocessing module (2) for performing data fusion based on each of the filtered biometric data, finding the optimal matching relationship between each of the filtered biometric data during the fusion process based on a preset optimization algorithm, and fusing to obtain multimodal fusion biometric data; A feature extraction module (4) connected to the multimodal fusion module (3) for performing feature extraction on the multimodal fusion biometric data according to a preset feature extraction algorithm, and dynamically adjusting the parameters of the feature extractor in the feature extraction algorithm according to each of the filtered environmental data during the feature extraction process to obtain a biometric template; A key generation module (5) connected to the feature extraction module (4) for generating a biometric dynamic key in real time according to the biometric template; A data encryption module (6) connected to the key generation module (5) for dynamically dividing the original data into multiple data shards according to the biometric dynamic key, and performing data encryption on each of the data shards using a lattice-based quantum-resistant algorithm to obtain encrypted shards; A shard recombination module (7) connected to the data encryption module (6) for providing a quantum-resistant shard recombination strategy to enable the user to restore each of the encrypted shards to the original data when providing the correct biometric data and the dynamic key; 2. The quantum-resistant data encryption system based on biometrics according to claim 1, wherein: The preprocessing module (2) includes: An adaptive filtering unit (21) for suppressing random noise in each of the biometric data according to an adaptive median filtering algorithm to obtain first filtered data; A multi-resolution analysis unit (22) for decomposing the images in each of the biometric data according to a multi-resolution method, and then restoring the detailed information in the images according to a threshold denoising technique to obtain second filtered data; A deep denoising unit (23) for performing image denoising and enhancement processing on the images in each of the biometric data according to a pre-trained deep learning algorithm to obtain third filtered data. The first filtered data, the second filtered data, and the third filtered data constitute the filtered biometric data.

3. The quantum-resistant data encryption system based on biometrics according to claim 1, characterized in that: The multimodal fusion module (3) includes: A feature alignment unit (31) is configured to extract iris vector features and fingerprint vector features of the iris image and the fingerprint image according to the ResNet algorithm, and extract heart rate time-series features and blood oxygen time-series features of the heart rate and the blood oxygen according to the LSTM network algorithm. Furthermore, according to the principal component analysis algorithm, the heart rate time-series features, the blood oxygen time-series features, the iris vector features and the fingerprint vector features are unified to the same dimension; A feature standardization unit (32) is connected to the feature alignment unit (31) and is configured to perform normalization processing on the heart rate time-series features, the blood oxygen time-series features, the iris vector features and the fingerprint vector features according to the Z-score normalization method to obtain a plurality of normalized features, where the normalized features include a heart rate normalized feature, a blood oxygen normalized feature, an iris normalized feature and a fingerprint normalized feature; A fusion unit (33) is connected to the feature standardization unit (32) and is configured to respectively configure initial weights for the normalized features, introduce an attention mechanism to perceive the feature quality of the normalized features, and dynamically adjust the corresponding initial weights according to the feature quality to obtain adjusted weights. Finally, the multimodal fusion biometric data is calculated according to the normalized features and the corresponding adjusted weights.

4. The quantum-resistant data encryption system based on biometrics according to claim 3, wherein: The calculation formula of the multimodal fusion biometric data is configured as: ； Among them, represents the multimodal fusion biological data, represents the normalized heart rate feature, represents the normalized blood oxygen feature, represents the normalized iris feature, represents the normalized fingerprint feature, represents the adjustment weight of the heart rate feature, represents the adjustment weight of the blood oxygen feature, represents the adjustment weight of the iris feature, represents the adjustment weight of the fingerprint feature, represents a preset offset constant, represents the Gamma function, represents the error function, represents the Bessel function of the first kind, represents the natural logarithm function, represents the upper limit of summation, represents the summation index variable, represents the summation denominator exponent parameter.

5. The quantum-resistant data encryption system based on biometrics according to claim 4, wherein: The multimodal fusion module (3) further includes an optimization matching unit (34) connected to the fusion unit (33), and is configured to search for the feature combination with the strongest correlation between the normalized features as the optimal matching relationship according to the maximum mutual information algorithm, and adjust the calculation formula of the multimodal fusion biometric data according to the normalized features and the corresponding adjusted weights in the optimal matching relationship to obtain the optimized multimodal fusion biometric data.

6. The quantum-resistant data encryption system based on biometrics according to claim 1, wherein: The feature extraction module (4) includes: A feature extraction unit (41) is configured to introduce a multi-layer residual convolution network and extract features of the multimodal fusion biometric data according to the multi-layer residual convolution network to obtain an intermediate feature tensor; A scoring modeling unit (42) is configured to construct an environment scoring model based on a shallow neural network, and input the light intensity, the temperature, the humidity and the background noise into the environment scoring model to obtain a parameter adjustment factor; A parameter adjustment unit (43) is connected to the scoring modeling unit (42) and is configured to dynamically adjust the convolution kernel coefficient of the feature extractor in the feature extraction algorithm according to the parameter adjustment factor to obtain an optimized extraction algorithm model, and the optimized extraction algorithm model is used to output an optimized feature tensor; A template generation unit (44) is connected to the parameter adjustment unit (43) and the feature extraction unit (41) and is configured to generate the biometric template according to the optimized feature tensor.

7. The quantum-resistant data encryption system based on biometrics according to claim 6, characterized in that: The calculation formula of the parameter adjustment factor is configured as: ； Among them, represents the parameter adjustment factor, and respectively represent two weight matrices of the environmental scoring model, and respectively represent two bias terms of the environmental scoring model, respectively represent the light intensity, the temperature, the humidity and the background noise, represents an activation function, represents the hyperbolic tangent function; The convolution kernel coefficient adjustment formula is configured as: ； Among them, represents the adjusted convolution kernel coefficients, represents the convolution kernel parameters before adjustment, represents the number of parameters to be adjusted, represents the th parameter adjustment factor.

8. The quantum-resistant data encryption system based on biometrics according to claim 1, characterized in that: The filtered environmental data further includes a sensor angle, a physical occlusion amount and a motion jitter amount. Then, the data encryption module (6) includes: A perturbation generation unit (61) generates a perturbation vector based on the sensor angle, the physical occlusion amount, and the motion jitter amount, and the perturbation vector is used to control the parameter setting of the data encryption process; A function generation unit (62) is connected to the perturbation generation unit (61) and is configured to generate a perturbation adjustment function according to the perturbation vector; A data encryption unit (63) is connected to the function generation unit (62) and is configured to dynamically adjust the number of shards, the number of encryption rounds, and the lattice password parameter configuration in the data encryption process according to the perturbation encryption function.

9. The quantum-resistant data encryption system based on biometrics according to claim 8, wherein: The perturbation adjustment function is configured as: ； Among them, represents the disturbance adjustment function, represents the disturbance vector, and represent two adjustable initial constants, respectively represent the sensor angle, the physical occlusion amount, and the motion jitter amount after normalization, represents three preset weight coefficients.

10. A quantum-resistant data encryption method based on biometrics, applied to the quantum-resistant data encryption system described in any one of claims 1-9, characterized in that, Including: Step S1, a data acquisition module (1) real-time collects a plurality of biometric data of a user and a plurality of environmental data in the environment where the user is located; Step S2, a preprocessing module (2) respectively performs data preprocessing on each of the biometric data and the environmental data according to a preset preprocessing algorithm to obtain processed filtered biometric data and filtered environmental data, where the filtered biometric data includes iris images, fingerprint images, heart rate, and blood oxygen, and the filtered environmental data includes light intensity, temperature, humidity, and background noise; Step S3, a multimodal fusion module (3) performs data fusion according to each of the filtered biometric data, and searches for an optimal matching relationship between each of the filtered biometric data based on a preset optimization algorithm during the fusion process, and fuses to obtain multimodal fusion biometric data; Step S4, a feature extraction module (4) performs feature extraction on the multimodal fusion biometric data according to a preset feature extraction algorithm, and dynamically adjusts the parameters of a feature extractor in the feature extraction algorithm according to each of the filtered environmental data during the feature extraction process to obtain a biometric template; Step S5, a key generation module (5) real-time generates a biometric dynamic key according to the biometric template; a data encryption module (6) dynamically divides the original data into a plurality of data shards according to the biometric dynamic key, and performs data encryption on each of the data shards using a lattice password-based quantum-resistant algorithm to obtain encrypted shards; Step S6, a shard recombination module (7) provides a quantum-resistant shard recombination strategy, so that when the user provides correct respective biometric data and the dynamic key, each of the encrypted shards is restored to the original data.

Images