A lawyer identity management and verification system based on human body feature information

Abstract

The present invention discloses a lawyer identity management verification system based on human characteristic information, which relates to the technical field of identity verification. The system includes synchronously acquiring the 3D facial image and electroencephalogram signal of a lawyer, extracting features from the 3D facial image to generate a facial feature vector, and performing a living body verification on the electroencephalogram signal. When the verification passes, a desensitized biometric data packet containing the facial feature vector and the practice number is output; an electronic signature audit chain is generated based on a dynamic permission credential, the electronic signature audit chain is associated with a blockchain storage credential and then written into the blockchain, and the feature extraction parameters are updated based on federated learning and the quantum encryption key is synchronized. Through the sub-microsecond-level multimodal synchronization method and spatio-temporal correlation living body verification, the present invention significantly reduces the synchronization error of biometric acquisition and greatly enhances the defense ability against high-precision counterfeiting attacks.

Description

Technical Field

[0001] The present invention relates to the technical field of identity authentication, and in particular to a lawyer identity management verification system based on human characteristic information. Background Art

[0002] In recent years, more and more identity authentication systems have begun to use multiple body characteristics such as face and brain waves simultaneously, which can better ensure security. In existing methods, facial feature extraction based on 3D point clouds combined with electroencephalogram (EEG) liveness detection has become the mainstream solution, and attempts have been made to combine blockchain technology for distributed storage. Dynamic watermark embedding and quantum encryption technology have been introduced into the field of biometric protection to address model stealing attacks and quantum computing threats. It should be noted that there are still significant deficiencies in the traditional technology in terms of the synchronization accuracy of multi-modal feature synchronous acquisition, the robustness of dynamic anti-counterfeiting marks, and the anti-quantum attack storage architecture.

[0003] The deficiencies of traditional technology mainly lie in the insufficient clock synchronization accuracy (usually at the millisecond level) between multi-modal biometric acquisition devices, resulting in inaccurate spatio-temporal correlation features in liveness detection, making it difficult to resist dual counterfeiting attacks of high-precision 3D masks and pre-recorded EEG signals. In traditional static watermark embedding strategies, it is easy to be cracked by reverse engineering, lacking a dynamic binding mechanism with practice information and unable to meet the traceability requirements in cross-border judicial cooperation. Existing blockchain storage solutions mostly adopt fixed redundant sharding strategies, which are difficult to dynamically adapt to network load fluctuations, thus unable to build a quantum-secure encryption system and facing the risk of data leakage under future quantum computing attacks. Summary of the Invention

[0004] In view of the above existing problems, the present invention is proposed.

[0005] Therefore, the present invention provides a lawyer identity management verification system based on human characteristic information to solve the problems of insufficient synchronization and anti-counterfeiting verification of multi-modal features and weak anti-quantum storage in lawyer identity verification.

[0006] To solve the above technical problems, the present invention provides the following technical solutions:

[0007] The present invention provides a lawyer identity management verification system based on human characteristic information, which includes a feature acquisition and verification module that synchronously acquires the 3D facial image and electroencephalogram signal of a lawyer, extracts features from the 3D facial image to generate a facial feature vector, and conducts a living body verification on the electroencephalogram signal. When the verification passes, a desensitized biometric data packet containing the facial feature vector and the practice number is output; a watermark encryption module that extracts the practice number from the desensitized biometric data packet, generates a dynamic watermark based on a chaotic encryption algorithm, and embeds the dynamic watermark into the facial feature vector to output a watermark-bearing enhanced feature vector; a distributed storage and indexing module that conducts quantum encryption on the enhanced feature vector, divides it into shards and stores them distributedly in blockchain nodes, constructs a two-layer index structure of hash and semantics, and generates a blockchain storage certificate; a verification and authorization module that, when a lawyer initiates a verification request, repeats the operations of the feature acquisition and verification module to the watermark encryption module to generate a temporary verification vector, matches the database record through the two-layer index structure, verifies the watermark consistency between the temporary verification vector and the stored enhanced feature vector, and generates a dynamic permission certificate; an audit and synchronization module that generates an electronic signature audit chain based on the dynamic permission certificate, associates the electronic signature audit chain with the blockchain storage certificate and writes it into the blockchain, and updates the feature extraction parameters and synchronizes the quantum encryption key based on federated learning.

[0008] As a preferred solution of the lawyer identity management verification system based on human characteristic information according to the present invention, in the synchronous acquisition operation, the acquisition devices of the 3D facial image and the electroencephalogram signal are clock-synchronized through an FPGA controller;

[0009] The feature extraction refers to using a PointNet++ dynamic graph convolutional network and combining a self-attention mechanism to extract the facial geometric feature vector of the eye socket, nose bridge, and lip region;

[0010] The electroencephalogram signal acquisition includes the composite features of alpha waves, beta waves, and PPG heart rate signals.

[0011] As a preferred solution of the lawyer identity management verification system based on human characteristic information according to the present invention, for the living body verification, the specific steps are as follows:

[0012] Calculate the power spectral density ratio of alpha waves and beta waves in the electroencephalogram signal;

[0013] Detect the micro-action fluctuation frequency of the 3D point cloud in the 3D facial image in the time series;

[0014] Set the electroencephalogram ratio threshold and the standard range of micro-action frequency;

[0015] When the power spectral density ratio exceeds the electroencephalogram ratio threshold and the micro-action frequency remains within the standard range of micro-action frequency, it is determined that the biological living body passes the verification.

[0016] As a preferred solution of the lawyer identity management verification system based on human characteristic information according to the present invention, wherein: the construction of the double-layer index structure of hash and semantics is specifically carried out as follows.

[0017] Use a collision-resistant hash algorithm to generate the hash fingerprint of the facial feature vector, and extract the entity semantic features in the practice number through a natural language processing model.

[0018] Combine the hash fingerprint and the entity semantic features to generate a composite index key value, and write it into the blockchain smart contract.

[0019] As a preferred solution of the lawyer identity management verification system based on human characteristic information according to the present invention, wherein: the generation of the dynamic permission certificate is specifically carried out as follows.

[0020] When initiating a verification request, the verification server synchronously collects 3D point cloud and electroencephalogram signals, performs liveness detection through a spatio-temporal convolutional network, and generates an enhanced verification vector embedded with a dynamic watermark.

[0021] Calculate the hash value based on the enhanced verification vector and compare it with the Merkle root stored in the blockchain. At the same time, perform double verification through semantic fingerprint matching and dynamic watermark consistency verification.

[0022] Encapsulate the verification result and sign it in a trusted execution environment, and authorize the generation of a dynamic permission certificate with a limited validity period.

[0023] As a preferred solution of the lawyer identity management verification system based on human characteristic information according to the present invention, wherein: the generation of the dynamic watermark based on the chaotic encryption algorithm is specifically carried out as follows.

[0024] Use the hash value of the practice number as the chaotic encryption seed, and generate an unpredictable watermark sequence through iterative operations.

[0025] Dynamically adjust the embedding depth of the watermark sequence based on the numerical distribution of the facial feature vector to form a dynamic watermark.

[0026] As a preferred solution of the lawyer identity management verification system based on human characteristic information according to the present invention, wherein: the distributed storage of the sharded data to the blockchain nodes refers to adopting an adaptive redundancy strategy, and dynamically calculating and configuring the redundancy parameters of the data shards based on the number of online nodes and the load status of the current blockchain network through the blockchain smart contract.

[0027] As a preferred solution of the lawyer identity management verification system based on human feature information according to the present invention, wherein: the federated learning to update the feature extraction parameters refers to aggregating the gradient update values of the feature extraction parameters based on the multi-party secure computing protocol, dynamically optimizing the facial feature extraction model in combination with historical verification data, and adaptively adjusting the sensitivity and coverage range of the feature extraction area.

[0028] As a preferred solution of the lawyer identity management verification system based on human feature information according to the present invention, wherein: the synchronization of quantum encryption keys means that the main path transmits new keys through an anti-quantum attack algorithm, the backup path ensures the recoverability of the keys through a key sharding trusteeship service, and a transition period for alternating new and old keys is set.

[0029] As a preferred solution of the lawyer identity management verification system based on human feature information according to the present invention, wherein: the electronic signature audit chain includes three levels of verification elements, which respectively record the fingerprint of the verification terminal device, the lawyer practice area code, and the permission operation timestamp, and each level of element uses an asymmetric encryption algorithm for digital signature.

[0030] The beneficial effects of the present invention are as follows: through the sub-microsecond multi-modal synchronization method and spatio-temporal correlation live body verification, the present invention significantly reduces the synchronization error of biometric data collection and greatly enhances the defense ability against high-precision counterfeiting attacks; combined with chaotic dynamic watermarking and anti-quantum storage architecture, while ensuring the watermark extraction accuracy, a quantum-secure encryption system is constructed to effectively resist the risk of data leakage; based on federated learning optimization and multi-level encryption audit chain, the cross-domain verification accuracy is significantly improved and high-precision operation traceability is achieved; by adopting adaptive feature fusion and deep reinforcement decision optimization, the verification stability in complex environments is significantly enhanced and noise interference is reduced, thus forming a complete protection chain from collection to storage. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for the description of the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present invention, and those of ordinary skill in the art can obtain other drawings without creative efforts based on these drawings.

[0032] Figure 1 It is a schematic diagram of the lawyer identity management verification system based on human feature information in Embodiment 1.

[0033] Figure 2 It is a schematic diagram of the feature collection and verification module in Embodiment 1.

[0034] Figure 3 It is a schematic diagram of the watermark encryption module in Embodiment 1.

[0035] Figure 4 This is a schematic diagram of the shared storage index module in Example 1. DETAILED DESCRIPTION

[0036] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the accompanying drawings.

[0037] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0038] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0039] Example 1, reference Figures 1 to 4 , which is the first embodiment of the present invention, provides a lawyer identity management and verification system based on human feature information, including the following steps:

[0040] The feature verification module synchronously obtains the lawyer's 3D facial image and EEG signal, extracts features from the 3D facial image to generate a facial feature vector, performs liveness verification on the EEG signal, and outputs a desensitized biometric data packet containing the facial feature vector and practice number when the verification is passed.

[0041] Specifically, the following steps are included:

[0042] In the synchronous acquisition operation, the acquisition devices of 3D facial images and EEG signals are clock synchronized through the FPGA controller.

[0043] Specifically, the structured light 3D camera is linked with a high-density dry electrode EEG cap to output a 3D point cloud and EEG signal stream with time stamp alignment;

[0044] Specifically, the FPGA controller generates a synchronization signal with microsecond accuracy, which is expressed as:

[0045] ;

[0046] In the formula, Represents the synchronous pulse signal in the time domain, represents the time variable, is the pulse period, , is a sequence of integers, represents a positive integer, such as , is the Dirac impulse function, which directly controls the sampling clocks of the 3D camera and the EEG cap through the hardware trigger pin;

[0047] Attach a global timestamp to the acquired data, denoted as,

[0048] ;

[0049] In the formula, represents the th timestamp, is the sampling sequence index, represents the initial time reference, that is, the time point of the 0th sampling, is the sampling time interval, (corresponding to a 256kHz clock);

[0050] Among them, the pulse period in the synchronization pulse signal, together with the timestamp interval , constitutes the time reference for multi-device synchronization. The (seconds) in the global timestamp corresponds to the 256Hz sampling rate of the EEG device, and needs to be aligned with the synchronization signal and the sampling moment of the 3D camera.

[0051] Calibrate the clock deviation between devices through a linear regression model to minimize the timestamp deviation:

[0052] ;

[0053] In the formula, represents that by optimizing the parameters and , the objective function reaches the minimum value, is the clock slope parameter, is the clock offset parameter, is the total number of timestamp data pairs, is the index of the timestamp data pair, represents the timestamp recorded by the 3D camera at the th sampling, represents the timestamp recorded by the EEG cap at the th sampling;

[0054] Output: Ensure the 3D point cloud and EEG signal stream with timestamp alignment error ;

[0055] Among them, the timestamp is used as the global reference for generating and ;

[0056] Synchronous pulse signal Ensure strict synchronization of device sampling and provide a high-quality data basis for linear regression correction.

[0057] Preferably, a hardware-level time synchronization system based on a programmable logic device achieves sub-millisecond alignment of multi-modal sensor data acquisition through pulse signal triggering, effectively solving the clock drift problem between heterogeneous devices.

[0058] Feature extraction refers to using the PointNet++ dynamic graph convolutional network and combining the self-attention mechanism to extract facial geometric feature vectors in the eye socket, nose bridge, and lip regions;

[0059] Specifically, the synchronized 3D point cloud is input into the PointNet++ dynamic graph convolutional network, and the self-attention mechanism is combined to extract facial geometric feature vectors (256-dimensional normalized feature vectors) in the eye socket, nose bridge, and lip regions;

[0060] Specifically, for the 3D point cloud, the local graph radius Is adaptively adjusted according to the point density;

[0061] Among them, the 3D point cloud is represented as,

[0062] ;

[0063] In the formula, Is the point cloud set, Represents the th 3D point, Represents the coordinate space of the 3D point, and the coordinates of each point are real numbers, Represents the total number of points in the point cloud set, Is the point cloud index;

[0064] Input the point cloud and construct the local dynamic graph as,

[0065] ;

[0066] In the formula, Represents the local graph of the th point, Represents the points in the point cloud set whose distance from the center point Is less than the radius , Represents the point And the center point Euclidean distance;

[0067] The layer feature update formula is as follows:

[0068] ;

[0069] In the formula, represents the feature vector of the -th point in the -th layer, represents the non-linear activation function, represents the summation operation on all neighboring points in the local graph , represents the relative position vector, represents the self-attention weight calculated based on the features of point and point , represents the feature vector of the -th point in the -th layer, represents the feature vector of the -th point in the -th layer;

[0070] Output: 256-dimensional normalized feature vector ;

[0071] Verify the 256-dimensional normalized feature vector with the electroencephalogram signal input in vivo.

[0072] Among them, the electroencephalogram signal acquisition includes the composite features of alpha waves, beta waves, and PPG heart rate signals.

[0073] It should be noted that although both "local graph " and "local graph of the -th point " represent local graphs, they correspond to different contexts and indexing methods respectively. Specifically, is used to describe the local graph of the -th point in the point cloud, where is the index of a specific point in the point cloud set. While is the local graph used in the feature update formula, representing the local graph of the -th point, where is the index of a specific point in the feature update process. Therefore, and differ in the indices of the points they correspond to. is used to describe a point in the point cloud, while is used to describe a point in the feature update process.

[0074] Index , and , For different contexts respectively. Index , are used to describe points in a point cloud set, where represents the center point, represents the distance from the center point less than of the points. And the index , are used in the feature update formula, where represents the point whose feature is currently being updated, represents the local map in the neighboring points. Therefore, , are used to describe the points in the point cloud set and their local maps, while , are used to describe the points and their local maps during the feature update process. Although they are all used to describe the relationship between the local map and the points, their contexts and uses are different.

[0075] Furthermore, calculate the power spectral density ratio of alpha waves and beta waves in the EEG signal;

[0076] Specifically, preprocess the composite features of alpha waves, beta waves and PPG heart rate signals, including eliminating mains interference through a 50Hz power frequency notch filter and removing baseline drift and high-frequency noise using a 0.5 - 45Hz band-pass filter; then segment the denoised continuous signal into 4-second segments with 50% overlap between adjacent segments, apply the Hanning window function to each segment to suppress spectral leakage; use the Welch method to calculate the power spectral density (PSD) of each segment, convert the time-domain signal to the frequency-domain energy distribution through the fast Fourier transform (FFT), with a frequency resolution of 0.25Hz;

[0077] After obtaining the PSD curve, numerically integrate the power values in the alpha wave frequency band (8 - 13Hz) and beta wave frequency band (13 - 30Hz) respectively, and use the trapezoidal rule to accumulate the power density values corresponding to all frequency points within each frequency band; divide the alpha wave integration result by the beta wave integration result to obtain the power spectral density ratio, and if there are multiple segments of data, take the arithmetic mean of the ratios of each segment; output the quantization index through threshold determination (such as marking as abnormal fluctuation when the alpha / beta ratio > 2.0), and record the original PSD curve, integration interval boundary values and signal-to-noise ratio (SNR≥20dB) as the verifiable basis for the calculation process.

[0078] Detect the micro-motion fluctuation frequency of the 3D facial point cloud in the time series;

[0079] Specifically, based on a continuous 3D point cloud sequence aligned by a global timestamp, a stable feature point set of key facial regions (eye sockets, nose wings, corners of the mouth) is selected. The Iterative Closest Point (ICP) algorithm is used to register adjacent point clouds frame by frame and calculate the three-dimensional displacement vectors of each feature point. The displacement vector sequence is preprocessed, and the Savitzky-Golay filter (window length 15 frames, 3rd-order polynomial) is used to smooth high-frequency noise and retain the physiological micro-motion frequency band of 0.1 - 5 Hz, generating the displacement-time curve of each feature point. The displacement curves of all feature points in the same region are subjected to principal component analysis (PCA), and the first principal component is extracted as the representative motion signal of this region. The short-time Fourier transform (STFT, window length 2 seconds, Hanning window, 75% overlap) is applied to the principal component signal, and the main frequency component with the highest energy proportion in the time-frequency spectrum is calculated, and the invalid fluctuations with an amplitude lower than the baseline noise threshold (such as ±0.1 mm) are removed. The distribution of the main frequencies of all regions within the time window is statistically analyzed, and the frequency point with the highest occurrence frequency is taken as the micro-motion fluctuation frequency, and its compliance with the physiological living standard range (0.8 - 2.5 Hz) is verified through the chi-square test. At the same time, the displacement variance of each feature point (required to be < 0.5 mm²) and the signal-to-noise ratio (SNR ≥ 15 dB) are recorded as anti-interference ability indicators.

[0080] Set the electroencephalogram ratio threshold and the standard range of micro-motion frequencies;

[0081] Setting of the electroencephalogram ratio threshold: By collecting electroencephalogram signal samples of healthy subjects in the resting state (n ≥ 500 cases), calculating the distribution characteristics of the α / β power spectral density ratio, using the ROC curve analysis method to determine the optimal threshold boundary point (such as when α / β > 2.0, the abnormal probability reaches 95%), and verifying the threshold sensitivity (> 90%) and specificity (> 85%) in combination with clinical diagnosis results (such as positive cases of epileptic seizures, attention deficits, etc.). Finally, it is optimized through a double-blind test and written into the system configuration.

[0082] Determination of the standard range of micro-motion frequencies: Based on a 3D facial point cloud database (including 5000 sets of living and spoofing attack data), the main frequency distribution of physiological micro-motions (living body 0.8 - 2.5 Hz) is statistically analyzed, and the 3σ principle is used to delimit the boundary of the standard range (mean ± 1.5 times the standard deviation), and outliers caused by environmental interference (such as light changes, device jitter) are excluded through the chi-square test. Finally, in combination with the requirements of the ISO / IEC 30107-3 living detection standard, the frequency interval with a confidence level > 99% is solidified as the judgment rule.

[0083] When the power spectral density ratio exceeds the electroencephalogram ratio threshold and the micro-motion frequency remains within the standard range of micro-motion frequencies, it is determined that the biological living body passes the verification.

[0084] Generate an irreversible hash value for the facial feature vector passed by live verification and the practice license number through the blind signature algorithm, and encapsulate it as a desensitized biometric data block (ciphertext in CBOR format) after encryption using AES-256-GCM;

[0085] Write the desensitized biometric data block, the timestamp of the acquisition process, and the device fingerprint into the append-only log of the trusted execution environment (TEE) to generate a final desensitized biometric data packet with a Merkle tree digest, and output it to the blockchain storage node.

[0086] Preferably, a spatio-temporal alignment of three-dimensional visual signals and neuroelectrophysiological signals is achieved through a hardware trigger mechanism, breaking through the technical bottleneck of single-modal vulnerability to spoofing attacks in traditional biometrics. This architecture can be extended to high-security scenarios such as judicial authentication and financial transactions, and its synchronization accuracy ensures the dynamic feature correlation of live detection.

[0087] The watermark encryption module extracts the practice license number from the desensitized biometric data packet, generates a dynamic watermark based on the chaotic encryption algorithm, and embeds the dynamic watermark into the facial feature vector to output a strengthened feature vector with a watermark.

[0088] Specifically, it includes the following steps:

[0089] Decrypt the encrypted data block (desensitized biometric data block) in the final desensitized biometric data packet, restore the original data using the pre-shared key, and separate the practice license number of the lawyer and the facial feature vector from it;

[0090] Specifically, use the AES-256-GCM algorithm to decrypt the encrypted data block in combination with the pre-shared key to restore the plaintext data;

[0091] It should be noted that "plaintext data" refers to the original desensitized content restored by decryption, specifically including the lawyer's practice license number and the 256-dimensional facial feature vector passed by live verification , and the two have formed an irreversible associated data pair through blind signature hashing before encryption.

[0092] Extract the practice license number of the lawyer from the plaintext data and the 256-dimensional facial feature vector .

[0093] Convert the hash value of the practice license number into the initial parameter of the chaotic system, generate a disordered sequence through the chaotic algorithm, and convert it into a binary watermark;

[0094] Specifically, calculate the hash value of the practice license number and normalize it to the initial value within the range of ; ;

[0095] Generate a 256 - dimensional chaotic sequence by iterating the Logistic chaotic map ;

[0096] Binarize the chaotic sequence into a 256 - bit dynamic watermark , expressed as,

[0097] ;

[0098] In the formula, is the -th binary value of the dynamic watermark, representing the watermark bit generated after determination according to the chaotic sequence value , and is the index of the element in the chaotic sequence;

[0099] Output the dynamic watermark ;

[0100] Calculate the adaptive parameter according to the numerical distribution range of the facial feature vector, and embed the binary watermark into the corresponding feature through the eigenvalue offset strategy to generate an enhanced feature vector;

[0101] Specifically, input the facial feature vector and the dynamic watermark , calculate the dynamic range of the facial feature vector, and determine the quantization step , expressed as,

[0102] ;

[0103] Use the quantization index modulation (QIM) method to embed the dynamic watermark into the facial feature vector , expressed as,

[0104] ;

[0105] In the formula, represents the enhanced facial feature vector, represents the updated result of the -th eigenvalue after embedding the watermark.

[0106] Perform standardization processing on the enhanced feature vector, encapsulate the data block in combination with the practice number and device identifier, attach the integrity verification digest after encryption, and output the enhanced feature vector with watermark;

[0107] Specifically, perform L2 normalization processing on the feature vector to ensure that its value range is consistent with the original feature vector.

[0108] The feature vector , practice number The device identifier is encapsulated into a JSON format data block, and at the same time, the AES-256-GCM algorithm is used to encrypt the JSON data block to generate ciphertext.

[0109] Calculate the Merkle tree digest of the ciphertext, append it to the encrypted data block, and output the watermarked enhanced feature vector ciphertext data block.

[0110] Write the watermarked enhanced feature vector into the immutable log of the secure execution environment to generate a storage instruction to initiate the subsequent distributed secure storage process.

[0111] Specifically, write the ciphertext data block into the append-only log of the trusted execution environment (TEE) to generate a timestamped storage request instruction that includes the Merkle tree digest of the ciphertext data block.

[0112] Send the storage request instruction to the blockchain node to initiate the subsequent quantum encryption sharding process.

[0113] The sharding index module performs quantum encryption on the enhanced feature vector, divides it into shards and distributes them to the blockchain nodes for storage, constructs a two-layer index structure of hash and semantics, and generates a blockchain storage certificate.

[0114] Specifically, it includes the following steps:

[0115] Generate a quantum-resistant public key and private key based on the NIST post-quantum cryptography standard CRYSTALS-Kyber algorithm, and store the private key in the hardware security unit of the trusted execution environment (TEE);

[0116] Specifically, call the NIST post-quantum cryptography library to initialize the CRYSTALS-Kyber algorithm parameters (such as security level L3, corresponding to a 256-bit key length).

[0117] Execute the Kyber key generation function to generate the public key and the private key .

[0118] Write the private key into the hardware security unit (HSM) of the trusted execution environment (TEE), and set the access permission (only internal calls within the TEE).

[0119] Register the public key on the blockchain , and bind it to the lawyer's practice number to ensure the uniqueness and traceability of the key.

[0120] Generate a hash fingerprint of the facial feature vector using a collision-resistant hash algorithm, and extract the entity semantic features in the practice number through a natural language processing model;

[0121] Furthermore, the encrypted data blocks are divided into dynamic redundant shards according to the Reed-Solomon erasure code rules, and an adaptive redundancy strategy is adopted. Based on the number of online nodes and the load status of the current blockchain network, the redundant parameters of the data shards are dynamically calculated and configured through the blockchain smart contract;

[0122] Specifically, according to the size of the encrypted data block (such as 1MB) and the load of the current blockchain nodes, the optimal redundant parameters are calculated through the smart contract .

[0123] Using the Reed-Solomon erasure code algorithm, the encrypted data block is divided into original shards, and redundant shards are generated.

[0124] Attach metadata to each redundant shard, including the shard serial number , shard hash and semantic labels (such as "lawyer identity verification").

[0125] Dynamically adjust the redundant parameters: If the node load increases, the smart contract automatically increases the redundant parameters (such as from (8,5) to (10,6)) to ensure data availability.

[0126] Use quantum public key to encrypt the shard key and bind it to the AES-256-GCM encrypted data shard to form a "quantum key + AES data" hybrid encrypted shard resistant to quantum attacks;

[0127] Specifically, for each shard , generate a temporary AES-256-GCM key inside the TEE, and encrypt the shard data, denoted as,

[0128] ;

[0129] Use the quantum public key to encrypt the temporary key , generate a quantum encrypted shard, denoted as,

[0130] ;

[0131] Bind and as a hybrid encrypted shard, and attach shard metadata (such as shard serial number and hash).

[0132] Destroy the temporary key to ensure the security of the shard data.

[0133] Calculate the hash of the shard content (SHA3-512) and the structure hash (Merkle tree root), and write the hash chain into the blockchain state database to achieve precise retrieval;

[0134] Specifically, for each hybrid encryption shard , calculate the content hash .

[0135] Construct a Merkle tree from the content hashes of all shards to generate the structure hash.

[0136] Write the content hash and the structure hash into the blockchain state database, and bind them to the lawyer's practice number to support precise retrieval.

[0137] Extract data semantic tags through the BERT-NER model, generate structured semantic vectors, and then compress them into 128-bit semantic fingerprints through Locality-Sensitive Hashing (LSH);

[0138] Specifically, for each hybrid encryption shard, calculate the hash value of its content, and use the SHA3-512 algorithm to ensure the uniqueness and collision resistance of the hash.

[0139] Construct a Merkle tree from the content hash values of all shards, and generate the root hash as the structure hash.

[0140] Write the content hash and the structure hash into the blockchain state database, and bind them to the lawyer's practice number to support subsequent precise retrieval;

[0141] Associate the semantic fingerprints with the shard content hashes to construct a distributed inverted index based on semantic similarity to support fuzzy query and context retrieval;

[0142] Specifically, extract key semantic tags from the metadata of the original data block, such as "lawyer identity verification" or "biometric data".

[0143] Use a pre-trained semantic recognition model to extract entity information from the tags and generate high-dimensional structured semantic vectors.

[0144] Compress the high-dimensional semantic vectors into fixed-length binary fingerprints through the Locality-Sensitive Hashing algorithm for subsequent construction of semantic indexes.

[0145] Combine the hash fingerprints with the entity semantic features to generate composite index key values and write them into the blockchain smart contract to ensure the immutability and traceability of the index data;

[0146] Subsequently, store the encrypted shards distributedly according to the node reputation scores based on IPFS-Cluster. The node returns the shard storage receipt and signs to verify the data integrity, thus achieving the secure storage and efficient retrieval of data shards;

[0147] Specifically, associate the generated binary semantic fingerprints with the sharded content hash values to construct a distributed inverted index.

[0148] Write the index data into the semantic index database of the blockchain to support fuzzy queries and context retrieval based on semantic tags.

[0149] Ensure the synchronization of the hash index and the semantic index through a smart contract to avoid data inconsistency.

[0150] Aggregate the shard hash, semantic index root hash, and dynamic sharding policy metadata within the TEE, and use a quantum private key to sign and generate an immutable blockchain storage credential;

[0151] Specifically, within the trusted execution environment, aggregate the hash values of the shards, the root hash of the semantic index, and the metadata of the dynamic sharding policy.

[0152] Use a quantum private key to sign the aggregated data to generate an immutable digital signature.

[0153] Encapsulate the signed data into a blockchain storage credential, and append a timestamp and a device fingerprint to ensure the uniqueness and traceability of the credential.

[0154] Write the hash index (Merkle tree root) and the semantic index (LSH encoded cluster) as transaction content into the blockchain to achieve a strong consistency binding between the index and the sharded data;

[0155] Specifically, write the root hash of the hash index and the binary fingerprint of the semantic index as transaction content into the blockchain.

[0156] Register the storage credential on the blockchain and bind it to the lawyer's practice number to support subsequent verification and auditing.

[0157] Verify the consistency between the index and the sharded data through a smart contract to ensure the integrity and reliability of the data.

[0158] Periodically check the availability of the shards through the blockchain smart contract, trigger the adaptive adjustment of the redundancy parameters, and update the sharding policy metadata in the storage credential.

[0159] Specifically, periodically check the availability of the shards through the blockchain smart contract, including the online status of the nodes and the integrity of the shards.

[0160] If a shard is found to be missing or damaged, trigger the adaptive adjustment of the redundancy parameters and regenerate the redundant shards.

[0161] Update the sharding policy metadata in the storage credential and synchronize it to the blockchain to ensure the high availability and consistency of the data.

[0162] Preferably, based on the dynamic watermark generation mechanism of the chaotic system and the feature space adaptive modulation strategy, a verifiable invisible identifier is implanted in the biometric vector. This technology effectively prevents model stealing attacks, provides traceable anti-counterfeiting guarantee for cross-border legal document authentication, and at the same time maintains the discriminative performance of the original features.

[0163] The verification authorization module, when a lawyer initiates a verification request, repeats the feature collection and verification module to the watermark encryption module to generate a temporary verification vector, matches the database records through a two-layer retrieval structure, verifies the watermark consistency between the temporary verification vector and the stored enhanced feature vector, and generates a dynamic permission certificate.

[0164] Specifically, it includes the following steps:

[0165] When initiating a verification request, the verification server synchronously collects 3D point cloud and EEG signals, performs liveness detection through a spatio-temporal convolutional network, and generates an enhanced verification vector for dynamic watermark embedding.

[0166] Specifically, the lawyer initiates a verification request through the client, encapsulates the practice number and identity information, encrypts it and sends it to the verification server.

[0167] The verification server starts the synchronous collection of the 3D camera and EEG cap, generates a 3D point cloud and EEG signal stream with time stamp alignment, and inputs it into the PointNet++ network. Combining with the self-attention mechanism, a 256-dimensional normalized feature vector is extracted.

[0168] Analyze the α / β wave energy ratio of the EEG signal, PPG heart rate variability, and the micro-action time series correlation of the 3D point cloud. Output the liveness score through the spatio-temporal convolutional network, and screen the legal signals with a score ≥ 0.95 to generate a temporary verification vector.

[0169] Extract the practice number from the verification request, generate a 256-bit dynamic watermark based on the Logistic chaotic map, and embed it into the temporary verification vector to generate an enhanced verification vector.

[0170] Calculate the hash value based on the enhanced verification vector and compare it with the Merkle tree root stored in the blockchain. At the same time, through semantic fingerprint matching and dynamic watermark consistency verification, double verification is performed.

[0171] Specifically, calculate the SHA3-512 hash value of the enhanced verification vector, compare it with the hash index (Merkle tree root) in the blockchain storage node, and enter the semantic index matching after successful matching.

[0172] Use the BERT-NER model to extract the semantic labels of the enhanced verification vector, generate a structured semantic vector and compress it into a 128-bit semantic fingerprint through locality-sensitive hashing (LSH), and compare it with the semantic index in the blockchain storage node.

[0173] Obtain the matching enhanced feature vectors from the blockchain storage nodes, generate the original watermark based on the same chaotic encryption algorithm, compare them bit by bit with the embedded watermark in the enhanced verification vector, and verify that the consistency is ≥ 95%.

[0174] Preferably, adopt the key encapsulation mechanism to combine classical encryption with post-quantum cryptography, construct a distributed storage scheme suitable for the quantum computing era, and achieve long-term auditability and efficient retrieval capabilities of judicial data.

[0175] Encapsulate and sign the verification result within the trusted execution environment, and authorize the generation of dynamic permission credentials with a limited validity period.

[0176] Specifically, within the trusted execution environment (TEE), encapsulate the verification result, practice number, timestamp, and device fingerprint into a JSON data packet, and use the quantum private key to sign and generate dynamic permission credentials.

[0177] Return the dynamic permission credentials to the client and write them into the blockchain storage nodes, and the client performs permission control operations according to the content of the credentials.

[0178] The audit synchronization module generates an electronic signature audit chain based on the dynamic permission credentials, associates the electronic signature audit chain with the blockchain storage credentials and then writes them into the blockchain, updates the feature extraction parameters based on federated learning, and synchronizes the quantum encryption key.

[0179] Specifically, it includes the following steps:

[0180] Use the quantum private key to sign the dynamic permission credentials to generate tamper-proof electronic signature data.

[0181] Among them, the electronic signature audit chain contains three-level verification elements, which respectively record the device fingerprint of the verification terminal, the lawyer's practice area code, and the permission operation timestamp, and each level of element uses an asymmetric encryption algorithm for digital signature.

[0182] Bind the electronic signature data to the Merkle root hash of the blockchain storage credentials to generate an associated data block.

[0183] Use the quantum public key to encrypt the associated data block and write it into the blockchain through a smart contract to generate blockchain storage credentials.

[0184] Initialize the federated learning framework within the trusted execution environment (TEE), and aggregate the gradient update values of the feature extraction parameters of each participating party based on the multi-party secure computation (MPC) protocol.

[0185] Distribute the aggregated global feature extraction parameter update values to each participating party through the MPC protocol to update the local feature extraction model.

[0186] Aggregate the gradient update values of the feature extraction parameters based on the multi-party secure computing protocol, dynamically optimize the facial feature extraction model in combination with historical verification data, and adaptively adjust the sensitivity and coverage of the feature extraction area.

[0187] Specifically, within the TEE, generate a new quantum key pair based on the NIST post-quantum cryptography standard CRYSTALS-Kyber algorithm, and write the new private key into the Hardware Security Module (HSM).

[0188] Distribute the new quantum public key to each participant through the federated learning framework, and register the new public key on the blockchain, marking the old public key as invalid.

[0189] Adopt a dual-path transmission mechanism to achieve quantum key synchronization: the main path transmits the new key through a quantum-resistant attack algorithm, the backup path ensures the recoverability of the key through a key sharding trusteeship service, and set an alternating transition period for the old and new keys to ensure the continuity and security of key updates.

[0190] Preferably, based on the optimization mechanism of secure multi-party computing, collaborative training of cross-domain biometric databases is achieved under the premise of protecting privacy. Break through the limitations of data silos and dynamically improve the generalization ability of the cross-border lawyer identity verification system.

[0191] In summary, the present invention significantly reduces the synchronization error of biometric acquisition and greatly enhances the defense ability against high-precision counterfeiting attacks through the sub-microsecond multi-modal synchronization method and spatio-temporal correlation live verification; combines chaotic dynamic watermarking and quantum-resistant storage architecture to build a quantum-secure encryption system while ensuring the watermark extraction accuracy, effectively resisting the risk of data leakage; based on federated learning optimization and multi-level encrypted audit chain, significantly improves the cross-domain verification accuracy and realizes high-precision operation traceability; adopts adaptive feature fusion and deep reinforcement decision optimization to significantly enhance the verification stability in complex environments and reduce noise interference, thus forming a complete protection chain from acquisition to storage.

[0192] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention, and they should all be covered within the scope of the claims of the present invention.

Claims

1. A lawyer identity management verification system based on human body feature information, characterized in that: include, The feature acquisition module synchronously acquires the lawyer's 3D facial image and EEG signal, extracts features from the 3D facial image to generate a facial feature vector, performs liveness verification on the EEG signal, and outputs a desensitized biometric data packet containing the facial feature vector and practice number when the verification is passed; The watermark encryption module extracts the practice number from the desensitized biometric data packet, generates a dynamic watermark based on a chaotic encryption algorithm, embeds the dynamic watermark into the facial feature vector, and outputs a watermarked enhanced feature vector; The distributed storage index module performs quantum encryption on the enhanced feature vector, divides it into shards and stores it in distributed form on blockchain nodes, builds a double-layer index structure of hash and semantics, and generates blockchain storage credentials; Verification and authorization module, when the lawyer initiates a verification request, repeats the feature collection module to the watermark encryption module to generate a temporary verification vector, matches the database record through the double-layer index structure, verifies the consistency of the temporary verification vector with the watermark of the stored enhanced feature vector, and generates a dynamic permission certificate. The specific steps are as follows: When initiating a verification request, the verification server synchronously collects 3D point clouds and EEG signals, performs liveness detection through a spatiotemporal convolutional network, and generates an enhanced verification vector embedded with a dynamic watermark; The hash value is calculated based on the enhanced verification vector and compared with the Merkle tree root stored in the blockchain. At the same time, double verification is performed through semantic fingerprint matching and dynamic watermark consistency verification; The double verification process is as follows: Calculate the SHA3-512 hash value of the enhanced verification vector, compare it with the hash index in the blockchain storage node, and enter the semantic index matching after a successful match; Use the BERT-NER model to extract the semantic labels of the enhanced verification vector, generate a structured semantic vector and compress it into a semantic fingerprint through local sensitive hashing, and compare it with the semantic index in the blockchain storage node; Obtain the matching enhanced feature vector from the blockchain storage node, generate the original watermark based on the same chaotic encryption algorithm, and compare it bit by bit with the embedded watermark in the enhanced verification vector to verify consistency; Encapsulate and sign the verification results in a trusted execution environment, and authorize the generation of dynamic permission credentials with a limited validity period; The audit synchronization module generates an electronic signature audit chain based on dynamic permission credentials, associates the electronic signature audit chain with the blockchain storage credentials and writes it into the blockchain, updates feature extraction parameters based on federated learning, and synchronizes quantum encryption keys; The federated learning to update feature extraction parameters refers to aggregating feature extraction parameter gradient update values based on a multi-party secure computing protocol, dynamically optimizing the facial feature extraction model in combination with historical verification data, and adaptively adjusting the sensitivity and coverage of the feature extraction area.

2. The lawyer identity management verification system based on human body feature information according to claim 1, wherein: In the synchronous acquisition operation, the acquisition devices of 3D facial images and EEG signals are clock synchronized through the FPGA controller; The feature extraction refers to using PointNet++ dynamic graph convolutional network combined with self-attention mechanism to extract facial geometric feature vectors of eye sockets, nose bridge and lip area; The EEG signal acquisition includes composite features of α wave, β wave and PPG heart rate signal.

3. The lawyer identity management verification system based on human body feature information according to claim 2, characterized in that: The liveness verification has the following specific steps: Calculate the power spectral density ratio of alpha waves and beta waves in the EEG signal; Detect the micro-motion fluctuation frequency of the 3D point cloud in the 3D facial image over time series; Set the brain wave ratio threshold and the standard range of micro-motion frequency; When the power spectral density ratio exceeds the brain wave ratio threshold and the micro-motion frequency is maintained within the standard range of micro-motion frequency, it is determined that the biological living body passes the verification.

4. The lawyer identity management verification system based on human body feature information according to claim 1, characterized in that: The construction of the double-layer index structure of hash and semantics is as follows: Use a collision-resistant hash algorithm to generate the hash fingerprint of the facial feature vector, and extract the entity semantic features in the practice number through a natural language processing model; Combine the hash fingerprint and the entity semantic features to generate a composite index key value and write it into the blockchain smart contract.

5. The lawyer identity management verification system based on human body feature information according to claim 2, characterized in that: The generation of the dynamic watermark based on the chaotic encryption algorithm is as follows: Use the hash value of the practice number as the chaotic encryption seed, and generate an unpredictable watermark sequence through iterative operations; Dynamically adjust the embedding depth of the watermark sequence based on the numerical distribution of the facial feature vector to form a dynamic watermark.

6. The lawyer identity management verification system based on human body feature information according to claim 4, characterized in that: The distributed storage of the sharded data to the blockchain nodes means adopting an adaptive redundancy strategy, based on the number of online nodes and the load status of the current blockchain network, and dynamically calculating and configuring the redundancy parameters of the data shards through the blockchain smart contract.

7. The lawyer identity management verification system based on human body feature information according to claim 1, characterized in that: The synchronization of the quantum encryption key means that the main path transmits the new key through an anti-quantum attack algorithm, the backup path ensures the recoverability of the key through the key sharding trusteeship service, and sets an alternating transition period between the old and new keys.

8. The lawyer identity management verification system based on human body feature information according to claim 1, wherein: The electronic signature audit chain contains three levels of verification elements, which respectively record the fingerprint of the verification terminal device, the lawyer's practice area code, and the permission operation timestamp, and each level of element is digitally signed using an asymmetric encryption algorithm.

Images