An electronic signature generation and anti-counterfeiting system based on multi-source information fusion

Abstract

The present invention discloses an electronic signature generation and anti-counterfeiting system based on multi-source information fusion, including: a multi-source information acquisition module for synchronously obtaining user fingerprint / iris features, the hardware fingerprint of the terminal device, and the timestamp and geographical location information during operation; an information fusion processing module that uses a weighted feature fusion algorithm and blockchain encryption anchoring technology to encrypt and compress the multi-source information and generate a unique feature vector; an electronic signature generation module that generates a visual signature graph containing a dynamic digital watermark and a multi-layer nested timestamp based on the feature vector, and at the same time writes the hash value of the feature vector into the blockchain for evidence storage; an anti-counterfeiting verification module that verifies the authenticity of the signature through the dynamic decryption protocol of quantum key distribution and supports reverse feature matching verification based on blockchain traceability. The present invention breaks through the limitations of traditional single encryption modes through the cross-binding of multi-modal biometric features and hardware fingerprints, combined with quantum keys and blockchain sharding technology.

Description

Technical Field

[0001] The present invention belongs to the technical field of electronic signature, and particularly relates to an electronic signature generation and anti-counterfeiting system based on multi-source information fusion. Background Art

[0002] With the rapid development of information technology, electronic signatures are increasingly widely used in fields such as e-government, e-commerce, and e-finance. Traditional electronic signature technologies mainly rely on a single information source, such as digital certificates or simple biometric recognition. These methods have certain limitations in terms of security, reliability, and adaptability. For example, single biometric recognition may be affected by forgery or replay attacks, while simple digital certificates may lose security due to the leakage of private keys.

[0003] To address these challenges, multi-source information fusion technology has gradually become a research hotspot in the field of electronic signatures. Multi-source information fusion can combine various different types of information sources, such as biometrics, device fingerprints, and dynamic environmental information. By comprehensively analyzing and processing this information, the security and reliability of electronic signatures can be improved. At the same time, the rise of blockchain technology provides a new idea for the anti-counterfeiting verification of electronic signatures. Its characteristics of distributed storage, immutability, and traceability can effectively enhance the credibility and anti-tampering ability of electronic signatures. Summary of the Invention

[0004] In view of this, the present invention proposes an electronic signature generation and anti-counterfeiting system based on multi-source information fusion, which breaks through the limitations of traditional single encryption modes through the cross-binding of multi-modal biometrics and hardware fingerprints, combined with quantum key and blockchain sharding technology.

[0005] To achieve the above object, the present invention provides the following technical solutions:

[0006] An electronic signature generation and anti-counterfeiting system based on multi-source information fusion provided by the present invention includes:

[0007] A multi-source information collection module, including a user biometric recognition unit, a device fingerprint extraction unit, and a dynamic environment perception unit, for synchronously obtaining user fingerprint / iris features, terminal device hardware fingerprints, and timestamp and geographical location information during operation;

[0008] An information fusion processing module, which uses a weighted feature fusion algorithm and blockchain encryption anchoring technology to encrypt and compress multi-source information and generate a unique feature vector;

[0009] An electronic signature generation module, which generates a visual signature graph containing a dynamic digital watermark and a multi-layer nested timestamp based on the feature vector, and at the same time writes the feature vector hash value into the blockchain for evidence storage;

[0010] The anti-counterfeiting verification module authenticates the authenticity of the signature through the dynamic decryption protocol of quantum key distribution and supports reverse feature matching verification based on blockchain traceability.

[0011] Preferably, the multi-source information acquisition module includes:

[0012] The user biometric recognition unit: used to collect fingerprint line features, iris texture phase encoding, and voiceprint dynamic spectrum through a fingerprint sensor, iris scanner, and microphone array in parallel through multi-threading;

[0013] The device fingerprint extraction unit: used to extract CPU microcode features, baseband chip serial numbers, and GPU rendering fingerprints by parsing the hardware registers of the terminal device and generate a composite device identification code;

[0014] The dynamic environment perception unit: used to capture the three-dimensional geographical coordinates, UTC timestamp, and device movement trajectory features during operation in real time through a built-in GPS / Beidou dual-mode positioning chip, NTP network time synchronization module, and acceleration sensor;

[0015] The multi-source information acquisition module runs in isolation through a security sandbox, and the collected data is transmitted to the information fusion processing module after being signed by HMAC-SHA3.

[0016] Preferably, the information fusion processing module includes:

[0017] The data preprocessing unit is used to normalize the biometric data. Among them, the fingerprint image is denoised by Gabor filtering, and the iris features are compressed to a 256-dimensional vector through Daubechies wavelet transform;

[0018] The weighted feature fusion unit dynamically assigns weight coefficients to biometric features, device features, and environmental features respectively based on a preset environmental risk assessment model, and calculates the fusion features through the weighted Euclidean distance formula:

[0019] , where , , are the real-time feature values of biometric features, device features, and environmental features dynamically respectively, , , are the reference values of biometric features, device features, and environmental features dynamically respectively, , , are the maximum tolerances of each dimension of biometric features, device features, and environmental features dynamically respectively, , , are the weight coefficients corresponding to biometric features, device features, and environmental features dynamically respectively;

[0020] The blockchain encryption anchoring unit is used to encrypt the fusion eigenvalue through the national cryptographic SM4 algorithm. After generating a 256-bit feature vector, its hash value is written into the Merkle tree of the Hyperledger Fabric consortium chain, and a timestamp and digital certificates of participating nodes are generated for each signature transaction.

[0021] Preferably, the electronic signature generation module includes:

[0022] The watermark embedding unit is used to generate a pseudo-random sequence by chaotic mapping of the feature vector, and embed a digital watermark in the intermediate frequency component of the signature image using discrete cosine transform. The embedding strength is adaptively adjusted based on network latency.

[0023] The timestamp generation unit is used to call the API of the National Time Service Center to obtain the Beidou atomic clock time, trigger the Ethereum smart contract to generate an on-chain time voucher containing the hash of the previous block, perform SHA-256 hash operation on the content of the signature file, encrypt it after splicing with the user's biometric snapshot to generate a timeliness voucher, and finally generate a multi-layer nested timestamp using the Beidou atomic clock time, on-chain time voucher, and timeliness voucher.

[0024] The signature generation unit is used to generate a visual signature graphic based on the digital watermark and multi-layer nested timestamp, and at the same time write the hash value of the feature vector into the blockchain for evidence storage. Among them, the signature graphic is stored in SVG vector format, including a watermark layer, a timestamp ciphertext layer, and a visible pattern layer.

[0025] Preferably, during the process of embedding the digital watermark in the intermediate frequency component of the signature image, the adjustment of the embedding strength is based on a feedback control model for network latency environment perception. The model is expressed as follows: , where represents the embedding strength of the intermediate frequency component, represents the reference strength, represents the real-time network latency, represents the preset latency threshold, represents the maximum tolerable latency, represents the adjustment coefficient.

[0026] Preferably, the anti-counterfeiting verification module includes:

[0027] The quantum key distribution unit is used to generate a temporary session key according to the decoy state BB84 protocol, distribute it through a 1550nm wavelength single-mode optical fiber channel, and is equipped with an NTRU-743 post-quantum encryption backup channel. The key lifecycle is strongly bound to the verification session.

[0028] The reverse feature matching unit is used to extract the SM4 encrypted storage hash value from the blockchain, decrypt it, and perform similarity comparison using the bandwidth-constrained derivative dynamic time warping algorithm. The matching threshold is dynamically adjusted according to the risk scenario, and the integrity of the restoration process is proved by zk-STARK;

[0029] The abnormal behavior analysis unit is used to deploy a bidirectional LSTM network to calculate the risk scores of the three elements of frequency, geography, and device in real time. When the trigger level ≥ 3, a three-step verification process of biometric re-authentication, device remote proof, and manual review is started.

[0030] Preferably, the weighted feature fusion algorithm in the information fusion processing module is also implemented through a spatio-temporal convolutional feature fusion network, which includes:

[0031] Multi-scale temporal convolutional layer: Dilated causal convolution is used to obtain the millisecond-level dynamic changes of biometric data, and the dilation rate of the convolutional kernel increases exponentially, The formula is: ,

[0032] where, represents the output of the dilated causal convolution, represents the input signal at time t, is the convolutional kernel weight, is the bias term, and k is the original convolutional kernel size;

[0033] Spherical space convolutional layer: Used to convert geographical location information into three-dimensional spherical coordinates: ,

[0034] where, represents longitude, represents latitude, represents the three-dimensional spherical coordinates;

[0035] Hierarchical spherical convolutional kernels are generated based on HEALPix grid partitioning, and the convolutional weight is constrained by the spherical harmonic function basis: ,

[0036] where, is the level, used to represent the resolution of the grid. The higher the level, the finer the grid, represents the spherical convolutional weight, indicating the convolutional weight at the spherical position , represents the coefficient of the spherical harmonic function, represents the spherical harmonic function, is the multiplicity, used to represent the orthogonal basis on the sphere, is the preset maximum order;

[0037] Output spatial feature map , where represents a spatial feature map in matrix form with dimension , represents the set of real numbers;

[0038] Quantum attention fusion layer: used to map temporal features and spatial features to the quantum state space: ,

[0039] where represents the temporal feature vector, and respectively represent the quantum state representations of the temporal feature and the spatial feature, and respectively represent the coefficients of the temporal feature and the spatial feature in the quantum ground state, representing the projection of the feature on the corresponding ground state, and respectively represent the preset orthogonal ground states, used to construct the quantum state space;

[0040] Generate an entangled state through a controlled-swap gate: ,

[0041] where represents the generated entangled state, used to represent the quantum entanglement of the temporal feature and the spatial feature, and represent the two ground states of the control qubit, represents the tensor product operation of quantum states, used to construct a composite quantum system; perform projective measurement on the entangled state to calculate the attention weight matrix :

[0042] ,

[0043] where represents the attention weight matrix The element in the i-th row and j-th column of, representing the attention weight between the i-th element of the temporal feature and the j-th element of the spatial feature, represents the observation operator of the element in the i-th row and j-th column, used to measure the correlation degree between the temporal feature and the spatial feature in the entangled state, represents the expected value of the observation operator on the entangled state, representing the correlation strength between the temporal feature and the spatial feature at the element in the i-th row and j-th column;

[0044] Output the fused feature vector , represents the operation of expanding the matrix into a vector, represents the attention weight matrix and the Kronecker product of the temporal feature and the spatial feature The matrix multiplication represents the weighted fusion of features.

[0045] The present invention has at least achieved the following beneficial effects:

[0046] 1. By cross-binding multi-modal biometric features with hardware fingerprints and combining quantum key and blockchain sharding technologies, the present invention breaks through the limitations of traditional single encryption modes.

[0047] 2. Utilizing the frequency-domain modulation technology of dynamic digital watermark + three-layer timestamp nesting mechanism significantly enhances the defense ability against deepfake attacks.

[0048] Other advantages, objectives, and features of the present invention will be described in the subsequent specification and will be obvious to those skilled in the art to some extent, or those skilled in the art can obtain teachings from the practice of the present invention. The objectives and other advantages of the present invention can be achieved and obtained through the following specification. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] To make the objectives, technical solutions, and beneficial effects of the present invention clearer, the present invention provides the following drawings for description:

[0050] Figure 1 It is a schematic structural diagram of an electronic signature generation and anti-counterfeiting system based on multi-source information fusion in an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0051] The following describes the preferred embodiments of the present invention with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are only for the purpose of illustrating and explaining the present invention and are not used to limit the present invention.

[0052] An electronic signature generation and anti-counterfeiting system based on multi-source information fusion provided by the present invention, with reference to Figure 1 , includes:

[0053] A multi-source information acquisition module, including a user biometric recognition unit, a device fingerprint extraction unit, and a dynamic environment perception unit, for synchronously acquiring user fingerprint / iris features, terminal device hardware fingerprints, and timestamps and geographical location information during operation;

[0054] An information fusion processing module, which uses a weighted feature fusion algorithm and blockchain encryption anchoring technology to encrypt and compress multi-source information and generate a unique feature vector;

[0055] An electronic signature generation module, which generates a visual signature graph containing dynamic digital watermarks and multi-layer nested timestamps based on the feature vector, and at the same time writes the feature vector hash value into the blockchain for evidence storage;

[0056] The anti-counterfeiting verification module authenticates the authenticity of the seal through the dynamic decryption protocol of quantum key distribution and supports reverse feature matching verification based on blockchain traceability.

[0057] The working principle and beneficial effects of the above technical solution are as follows: The present invention synchronously obtains user biometric features, terminal device hardware fingerprints, and timestamp and geographical location information during operation through the multi-source information acquisition module, and uses the weighted feature fusion algorithm and blockchain encryption anchoring technology of the information fusion processing module to encrypt and compress the multi-source information to generate a unique feature vector. The electronic seal generation module creates a visual seal graphic containing a dynamic digital watermark and a multi-layer nested timestamp based on the feature vector, and writes the hash value of the feature vector into the blockchain for evidence storage. The anti-counterfeiting verification module uses the dynamic decryption protocol of quantum key distribution to verify the authenticity of the seal, and at the same time supports reverse feature matching verification based on blockchain traceability. The entire solution improves the security and credibility of the seal, ensures the integrity and immutability of the data, is easy to operate, adapts to various scenarios, and has high efficiency and flexibility.

[0058] In a preferred embodiment, the multi-source information acquisition module includes:

[0059] User biometric feature recognition unit: used to collect fingerprint line features, iris texture phase encoding, and voiceprint dynamic spectrum through a fingerprint sensor, iris scanner, and microphone array in parallel through multiple threads;

[0060] Device fingerprint extraction unit: used to extract CPU microcode features, baseband chip serial numbers, and GPU rendering fingerprints by parsing the hardware registers of the terminal device and generate a composite device identification code;

[0061] Dynamic environment perception unit: used to capture three-dimensional geographical coordinates, UTC timestamp, and device movement trajectory features during operation in real time through a built-in GPS / Beidou dual-mode positioning chip, NTP network time service module, and acceleration sensor;

[0062] The multi-source information acquisition module runs in isolation through a security sandbox, and the collected data is transmitted to the information fusion processing module after being signed by HMAC-SHA3.

[0063] The working principle and beneficial effects of the above technical solution are as follows: The multi-source information acquisition module comprehensively obtains user biometric features, device fingerprints, and dynamic environment information through various sensors and parsing technologies. In the secure sandbox, after being encrypted and signed, these data are transmitted to the next module, providing a reliable and rich data source for subsequent information fusion processing and laying the foundation for the generation and anti-counterfeiting of electronic signatures. The multi-dimensional biometric feature acquisition and device fingerprint extraction, combined with dynamic environment perception, make the signature generation based on rich and accurate information, which is difficult to forge. The isolated operation of the secure sandbox and the HMAC-SHA3 signature further ensure the secure transmission of data and prevent data leakage and tampering. The multi-threaded parallel acquisition of multiple biometric features improves the recognition accuracy, effectively distinguishes different users, and enhances the user identity authentication function of the signature. The dynamic environment perception unit provides spatio-temporal related data, making the signature carry the spatio-temporal information during operation, enhancing the spatio-temporal attributes of the signature, facilitating subsequent traceability and verification, and meeting the spatio-temporal authentication requirements in specific scenarios. The multi-threaded parallel acquisition and the secure sandbox mechanism improve the efficiency of data acquisition and transmission while ensuring data security, reduce the operation waiting time, and enhance the user experience.

[0064] In a preferred embodiment, the information fusion processing module includes:

[0065] A data preprocessing unit for normalizing biometric data. Among them, Gabor filtering is used to denoise fingerprint images, and iris features are compressed to 256-dimensional vectors through Daubechies wavelet transform;

[0066] A weighted feature fusion unit that dynamically assigns weight coefficients to biometric features, device features, and environmental features respectively based on a preset environmental risk assessment model, and calculates the fusion features through the weighted Euclidean distance formula:

[0067] , where , , are the dynamic real-time feature values of biometric features, device features, and environmental features respectively, , , are the benchmark values of biometric features, device features, and environmental features respectively, , , are the maximum tolerances of each dimension of biometric features, device features, and environmental features respectively, , , are the weight coefficients corresponding to biometric features, device features, and environmental features respectively;

[0068] The blockchain encryption anchoring unit is used to encrypt the fusion eigenvalue through the national cryptography SM4 algorithm. After generating a 256-bit feature vector, its hash value is written into the Merkle tree of the Hyperledger Fabric consortium blockchain, and a timestamp and digital certificates of participating nodes are generated for each signature transaction.

[0069] The working principle and beneficial effects of the above technical solution are as follows: The information fusion processing module normalizes the biometric data through the data preprocessing unit to remove noise and redundant information, improving the data quality. The weighted feature fusion unit dynamically assigns weight coefficients according to the environmental risk assessment model to highlight important features, improving the pertinence and effectiveness of feature fusion. The blockchain encryption anchoring unit encrypts the fusion eigenvalue using the national cryptography SM4 algorithm to ensure the confidentiality and integrity of the data. The generated 256-bit feature vector and its hash value are written into the Merkle tree of the consortium blockchain, and the trusted storage of the feature vector is realized by using the distributed storage and immutability characteristics of the blockchain. A timestamp and digital certificates of participating nodes are generated for each signature transaction to ensure the chronological order of signature transactions and the traceability of the identities of participating nodes, enhancing the security and credibility of the system.

[0070] In a preferred embodiment, the electronic signature generation module includes:

[0071] The watermark embedding unit is used to generate a pseudo-random sequence by chaotic mapping of the feature vector, and embed a digital watermark in the middle-frequency component of the signature image using discrete cosine transform. The embedding strength is adaptively adjusted based on network latency;

[0072] The timestamp generation unit is used to separately call the API of the National Time Service Center to obtain the Beidou atomic clock time, trigger the Ethereum smart contract to generate an on-chain time voucher containing the hash of the previous block, perform a SHA-256 hash operation on the content of the signature file, encrypt it after splicing with the user's biometric snapshot to generate a timeliness voucher, and finally generate a multi-layer nested timestamp using the Beidou atomic clock time, the on-chain time voucher, and the timeliness voucher;

[0073] The signature generation unit is used to generate a visual signature graph according to the digital watermark and the multi-layer nested timestamp, and at the same time write the hash value of the feature vector into the blockchain for storage. Among them, the signature graph is stored in SVG vector format and includes a watermark layer, a timestamp ciphertext layer, and a visible pattern layer.

[0074] The working principle and beneficial effects of the above technical solution are as follows: Through the collaborative work of the watermark embedding unit, timestamp generation unit, and signature generation unit, the electronic signature generation module realizes the secure generation and archiving of electronic signatures. The watermark embedding unit uses chaotic mapping and discrete cosine transform to embed digital watermarks in the signature image. The embedding strength is adaptively adjusted according to network latency, which not only ensures the robustness of the watermark but also improves the visual quality of the signature image. The timestamp generation unit integrates the time of the Beidou atomic clock, the on-chain time certificate generated by the Ethereum smart contract, and the SHA-256 hash operation result of the signature file content to generate a multi-layer nested timestamp, ensuring the time sequence and content integrity of the signature file, and enhancing the credibility and anti-tampering ability of the signature. The signature generation unit combines the digital watermark and the multi-layer nested timestamp to generate a visual signature graphic, which is stored in SVG vector format. It not only includes the watermark layer and the timestamp ciphertext layer but also the visible pattern layer, making the signature visually recognizable and containing rich anti-counterfeiting information. At the same time, the feature vector hash value is written into the blockchain for archiving, and the immutability of the blockchain is used to further ensure the security and traceability of the signature. Through the multi-unit collaboration of the electronic signature generation module, this technical solution integrates key information such as feature vectors and timestamps into the signature image, generating an electronic signature with high security and credibility. This technical solution improves the security of the signature, enhances the credibility of the signature, ensures the integrity and traceability of the signature file, and improves the visual quality and anti-counterfeiting ability of the signature image.

[0075] In a preferred embodiment, during the process of embedding a digital watermark in the intermediate frequency component of the signature image, the adjustment of the embedding strength is based on a feedback control model for network latency environment perception. The model is expressed as follows: The model is expressed as follows: , where represents the embedding strength of the intermediate frequency component, represents the reference strength, represents the real-time network latency, represents the preset latency threshold, represents the maximum tolerable latency, represents the adjustment coefficient.

[0076] In a preferred embodiment, the anti-counterfeiting verification module includes:

[0077] A quantum key distribution unit, which is used to generate a temporary session key according to the decoy state BB84 protocol and distribute it through a 1550nm wavelength single-mode fiber optic channel. It is equipped with an NTRU-743 post-quantum encryption backup channel, and the key lifecycle is strongly bound to the verification session;

[0078] The reverse feature matching unit is used to extract the SM4 encrypted deposit hash value from the blockchain. After decryption, the derivative dynamic time warping algorithm with bandwidth constraint is used for similarity comparison. The matching threshold is dynamically adjusted according to the risk scenario, and the integrity of the restoration process is proved by zk-STARK.

[0079] The abnormal behavior analysis unit is used to deploy a bidirectional LSTM network to calculate the risk scores of the three elements of frequency, geography, and device in real time. When the trigger level ≥ 3, a three-step verification process of biometric re-authentication, device remote proof, and manual review is started.

[0080] The working principle and beneficial effects of the above technical solution are as follows: The anti-counterfeiting verification module realizes the security verification and risk prevention and control of the electronic signature through the collaborative work of the quantum key distribution unit, the reverse feature matching unit, and the abnormal behavior analysis unit. The quantum key distribution unit uses the decoy-state BB84 protocol to generate a temporary session key and distributes it through a 1550nm wavelength single-mode fiber channel. At the same time, an NTRU-743 post-quantum encryption backup channel is equipped to ensure the security and reliability of key distribution. The key life cycle is strongly bound to the verification session, further enhancing the security of key management. The reverse feature matching unit extracts the SM4 encrypted deposit hash value from the blockchain. After decryption, the derivative dynamic time warping algorithm with bandwidth constraint is used for similarity comparison. The matching threshold is dynamically adjusted according to the risk scenario, ensuring the accuracy and adaptability of feature matching. The integrity of the restoration process is proved by zk-STARK, ensuring the transparency and credibility of the verification process. The abnormal behavior analysis unit deploys a bidirectional LSTM network to calculate the risk scores of the three elements of frequency, geography, and device in real time. When the trigger level ≥ 3, a three-step verification process of biometric re-authentication, device remote proof, and manual review is started, effectively preventing abnormal behaviors and potential risks, and improving the security and reliability of the system. Through the multi-unit collaboration of the anti-counterfeiting verification module, this technical solution uses means such as quantum key distribution, blockchain technology, dynamic time warping algorithm, and abnormal behavior analysis to realize the security verification and risk prevention and control of the electronic signature. This technical solution improves the security of signature verification, enhances the risk prevention and control ability of the system, ensures the transparency and credibility of the verification process, and realizes the effective monitoring and timely response to abnormal behaviors.

[0081] In a preferred embodiment, the weighted feature fusion algorithm in the information fusion processing module is also implemented through a spatio-temporal convolutional feature fusion network, which includes:

[0082] Multi-scale temporal convolutional layer: Use dilated causal convolution to obtain the millisecond-level dynamic changes of biometric data, and the dilation rate of the convolution kernel increases according to an exponential rule, The formula is: ,

[0083] Among them, represents the output of the dilated causal convolution, represents the input signal at time t, is the convolution kernel weight, is the bias term, and k is the original convolution kernel size;

[0084] Spherical space convolution layer: used to convert geographical location information into three-dimensional spherical coordinates: ,

[0085] Among them, represents longitude, represents latitude, represents the three-dimensional spherical coordinates;

[0086] Generate a hierarchical spherical convolution kernel based on HEALPix grid partitioning, and the convolution weight is constrained by the spherical harmonic function basis: ,

[0087] Among them, is the level, used to represent the resolution of the grid. The higher the level, the finer the grid, represents the spherical convolution weight, indicating the convolution weight at the spherical position , represents the coefficient of the spherical harmonic function, represents the spherical harmonic function, is the multiplicity, used to represent the orthogonal basis on the sphere, is the preset maximum order;

[0088] Output spatial feature map , where represents the spatial feature map in matrix form with dimension , represents the set of real numbers;

[0089] Quantum attention fusion layer: used to map time features and spatial features to the quantum state space: ,

[0090] Among them, represents the time feature vector, and respectively represent the quantum state representations of the time feature and the spatial feature, and respectively represent the coefficients of the time feature and the spatial feature in the quantum ground state, indicating the projection of the feature on the corresponding ground state, and respectively represent the preset orthogonal ground states, used to construct the quantum state space;

[0091] Generate an entangled state through a controlled swap gate: ,

[0092] Among them, represents the generated entangled state, which is used to represent the quantum entanglement of time characteristics and space characteristics, and represent the two ground states that control qubits, represents the tensor product operation of quantum states, which is used to construct a composite quantum system; perform projective measurement on the entangled state to calculate the attention weight matrix :

[0093] ,

[0094] Among them, represents the attention weight matrix The element in the i-th row and j-th column of represents the attention weight between the i-th element of the time feature and the j-th element of the space feature, represents the observation operator of the element in the i-th row and j-th column, which is used to measure the correlation degree between the time feature and the space feature in the entangled state, represents the expected value of the observation operator on the entangled state, which represents the correlation strength between the time feature and the space feature at the element in the i-th row and j-th column;

[0095] Output the fused feature vector , represents the operation of expanding the matrix into a vector, represents the attention weight matrix The Kronecker product with the time feature and the space feature The matrix multiplication of represents the weighted fusion of features.

[0096] The beneficial effects of the above technical solution are that the spatio-temporal convolutional feature fusion network includes a multi-scale time convolutional layer, a spatial feature extraction layer, and a spatio-temporal feature fusion layer. The multi-scale time convolutional layer uses dilated causal convolution to obtain the millisecond-level dynamic changes of biometric data, and the dilation rate of the convolutional kernel increases according to an exponential rule, which can capture the dynamic changes of biometric data at different time scales. The spatial feature extraction layer extracts spatial features from device fingerprints and geographical location information, uses convolutional operations to extract local spatial features, and reduces the feature dimension through pooling operations. The spatio-temporal feature fusion layer fuses the temporal features output by the multi-scale time convolutional layer and the spatial features output by the spatial feature extraction layer, uses a weighted feature fusion algorithm to assign different weights to different features, and generates a comprehensive feature vector. The entire technical solution deeply processes and fuses multi-source information through the spatio-temporal convolutional feature fusion network, improving data quality and feature expression ability. This technical solution improves data quality, enhances feature expression ability, improves system security and credibility, and realizes efficient data fusion and processing.

[0097] Finally, it should be noted that the above preferred embodiments are only used to illustrate the technical solutions of the present invention rather than to limit them. Although the present invention has been described in detail through the above preferred embodiments, those skilled in the art should understand that various changes can be made in form and details without departing from the scope defined by the claims of the present invention.

Claims

1. An electronic signature generation and anti-counterfeiting system based on multi-source information fusion, characterized in that Including: A multi-source information acquisition module, including a user biometric recognition unit, a device fingerprint extraction unit, and a dynamic environment perception unit, for synchronously acquiring user fingerprint / iris features, terminal device hardware fingerprints, and timestamps and geographical location information during operation; An information fusion processing module, which uses a weighted feature fusion algorithm and blockchain encryption anchoring technology to encrypt and compress multi-source information and generate a unique feature vector; An electronic signature generation module, which generates a visual signature graph containing a dynamic digital watermark and a multi-layer nested timestamp based on the feature vector, and writes the hash value of the feature vector into the blockchain for evidence storage; An anti-counterfeiting verification module, which verifies the authenticity of the signature through a dynamic decryption protocol of quantum key distribution, and supports reverse feature matching verification based on blockchain traceability; Among them, the electronic signature generation module includes: A watermark embedding unit, which is used to generate a pseudo-random sequence by chaotic mapping of the feature vector, and embeds a digital watermark in the intermediate frequency component of the signature image by discrete cosine transform, and the embedding strength is adaptively adjusted based on network latency; A timestamp generation unit, which is used to separately call the API of the National Time Service Center to obtain the Beidou atomic clock time, trigger an Ethereum smart contract to generate an on-chain time certificate containing the hash of the previous block, perform a SHA-256 hash operation on the content of the signature file, encrypt it after splicing with the user biometric snapshot to generate a timeliness certificate, and finally generate a multi-layer nested timestamp using the Beidou atomic clock time, the on-chain time certificate, and the timeliness certificate; A signature generation unit, which is used to generate a visual signature graph based on the digital watermark and the multi-layer nested timestamp, and write the hash value of the feature vector into the blockchain for evidence storage. Among them, the signature graph is stored in SVG vector format and includes a watermark layer, a timestamp ciphertext layer, and a visible pattern layer.

2. The electronic signature generation and anti-counterfeiting system based on multi-source information fusion according to claim 1, wherein The multi-source information acquisition module includes: A user biometric recognition unit: used to collect fingerprint line features, iris texture phase encoding, and voiceprint dynamic spectrum through a fingerprint sensor, an iris scanner, and a microphone array in parallel through multiple threads; A device fingerprint extraction unit: used to extract CPU microcode features, baseband chip serial numbers, and GPU rendering fingerprints by parsing the hardware registers of the terminal device, and generate a composite device identification code; A dynamic environment perception unit: used to capture three-dimensional geographical coordinates, UTC timestamp, and device movement trajectory features during operation through a built-in GPS / Beidou dual-mode positioning chip, an NTP network time service module, and an acceleration sensor; The multi-source information acquisition module runs in isolation through a security sandbox, and the collected data is transmitted to the information fusion processing module after being signed by HMAC-SHA3.

3. An electronic signature generation and anti-counterfeiting system based on multi-source information fusion according to claim 1, characterized in that, The information fusion processing module includes: A data preprocessing unit, which is used to perform normalization processing on biometric data. Among them, Gabor filtering is used to denoise fingerprint images, and iris features are compressed to 256-dimensional vectors through Daubechies wavelet transform; A weighted feature fusion unit, which dynamically assigns weight coefficients to biometric features, device features, and environmental features respectively based on a preset environmental risk assessment model, and calculates the fusion feature through a weighted Euclidean distance formula: , where , , are the real-time eigenvalue of the dynamic biometric feature, device feature, and environmental feature respectively, , , are the reference values of the dynamic biometric feature, device feature, and environmental feature respectively, , , are the maximum tolerances of each dimension of the dynamic biometric feature, device feature, and environmental feature respectively, , , are the corresponding weight coefficients of the dynamic biometric feature, device feature, and environmental feature respectively; The blockchain encryption anchoring unit is used to encrypt the fusion eigenvalue through the national cryptography SM4 algorithm. After generating a 256-bit feature vector, its hash value is written into the Merkle tree of the Hyperledger Fabric consortium chain, and a timestamp and digital certificates of participating nodes are generated for each signature transaction.

4. The electronic signature generation and anti-counterfeiting system based on multi-source information fusion according to claim 1, characterized in that, In the process of embedding digital watermark in the intermediate frequency component of the signature image, the adjustment of the embedding strength is based on a feedback control model for network latency environment perception, and the model is expressed as follows: , where represents the embedding strength of the intermediate frequency component, represents the reference strength, represents the real-time network latency, represents the preset latency threshold, represents the maximum tolerable latency, represents the adjustment coefficient.

5. The electronic signature generation and anti-counterfeiting system based on multi-source information fusion according to claim 1, characterized in that, The anti-counterfeiting verification module includes: The quantum key distribution unit is used to generate a temporary session key according to the decoy state BB84 protocol, distribute it through a 1550nm wavelength single-mode optical fiber channel, and is equipped with an NTRU-743 post-quantum encryption backup channel. The key life cycle is strongly bound to the verification session; The reverse feature matching unit is used to extract the SM4 encrypted deposit hash value from the blockchain. After decryption, the bandwidth-constrained derivative dynamic time warping algorithm is used for similarity comparison. The matching threshold is dynamically adjusted according to the risk scenario, and the integrity of the reduction process is proved by zk-STARK; The abnormal behavior analysis unit is used to deploy a bidirectional LSTM network to calculate the risk scores of the three elements of frequency, geography, and device in real time. When the trigger level ≥ 3, a three-step verification process of biometric re-authentication, device remote proof, and manual review is started.

6. The electronic signature generation and anti-counterfeiting system based on multi-source information fusion according to claim 1, characterized in that, The weighted feature fusion algorithm in the information fusion processing module is also implemented through a spatio-temporal convolutional feature fusion network, which includes: Multi-scale temporal convolutional layer: dilated causal convolution is used to obtain millisecond-level dynamic changes of biometric data, and the dilation rate of the convolutional kernel increases exponentially, The formula is: , Among them, represents the output of the dilated causal convolution, represents the input signal at time t, is the convolutional kernel weight, is the bias term, and k is the size of the original convolutional kernel; Spherical space convolution layer: used to convert geographical location information into three-dimensional spherical coordinates: , Among them, represents the longitude, represents the latitude, represents the three-dimensional spherical coordinates; Generate a hierarchical spherical convolution kernel based on HEALPix grid division, and the convolution weights Constrained by the spherical harmonic function basis: ， Among them, is the level, used to represent the resolution of the grid. The higher the level, the finer the grid, represents the spherical convolution weight, indicating the convolution weight at the spherical position ; represents the coefficient of the spherical harmonic function, represents the spherical harmonic function, is the multiplicity, used to represent the orthogonal basis on the sphere, is the preset maximum order; Output spatial feature map , where represents a spatial feature map in matrix form with a dimension of , and represents the set of real numbers; The quantum attention fusion layer: used to map time features and space features to the quantum state space; ， Among them, represents the time feature vector, and respectively represent the quantum state representations of the time feature and the spatial feature, and respectively represent the coefficients of the time feature and the spatial feature in the quantum ground state, indicating the projection of the feature on the corresponding ground state, and respectively represent the preset orthogonal ground states for constructing the quantum state space; Generating entangled states through controlled-swap gates: , Among them, represents the generated entangled state, which is used to represent the quantum entanglement of time characteristics and spatial characteristics, and represents the two ground states that control the quantum bits, represents the tensor product operation of quantum states, which is used to construct a composite quantum system; perform projective measurement on the entangled state to calculate the attention weight matrix : ， Among them, represents the attention weight matrix The element in the i-th row and j-th column in, representing the attention weight between the i-th element of the time feature and the j-th element of the spatial feature, represents the observation operator for the element in the i-th row and j-th column, used to measure the correlation degree between the time feature and the spatial feature in the entangled state, represents the expected value of the observation operator on the entangled state, representing the correlation strength between the time feature and the spatial feature at the element in the i-th row and j-th column; Output fused feature vector , represents the operation of expanding a matrix into a vector, represents the attention weight matrix and the Kronecker product of the temporal feature and the spatial feature matrix multiplication of, representing the weighted fusion of features.

Images