High and new technology data asset management method based on multi-feature fusion

Abstract

The invention discloses a high and new technology data asset management method based on multi-feature fusion, and the method comprises the steps: constructing a management framework based on actual demands; image data and text data related to the high and new technology data assets are obtained and preprocessed, and text feature vectors and image feature vectors are obtained after feature extraction is carried out on the image data and the text data; fusing the text feature vector and the image feature vector through a cross-modal attention mechanism to obtain an image-text feature fusion vector after local feature fusion, and obtaining a multi-modal fusion global feature by using a multi-head attention mechanism; and classifying the data according to the multi-modal fusion global features to form an asset directory of high and new technology data assets. According to the method, a multi-modal feature fusion model is established, cross-modal image-text feature fusion extraction is realized, respective information, common information and associated information of images and texts are mined, and text features are improved by using a multi-attention mechanism, so that the feature extraction model has higher accuracy and the asset management efficiency is improved.

Description

Technical Field

[0001] The present invention relates to data management, and specifically to a high-tech data asset management method based on multi-feature fusion. Background Art

[0002] High-tech assets usually refer to tangible or intangible assets owned in the field of high-tech. These assets often involve the latest scientific and technological inventions, innovative achievements, patented technologies, trademarks, software copyrights, etc. High-tech assets are technical resources with economic value formed by enterprises, research institutions or individuals in the process of scientific research, technological development and high-tech industrialization. Among them, patented technology, as an asset containing rich technical information, intellectual property protection, commercialization potential and transaction feasibility, is regarded as one of the important data assets. The purpose of high-tech asset management is to ensure the safety of high-tech assets, improve their utilization rate, promote their value-added, and support the strategic development of enterprises. Through effective asset management, enterprises can protect core competitiveness, promote technological innovation, achieve industrial upgrading, maintain data security, and promote knowledge sharing, thereby gaining an advantage in the fierce market competition.

[0003] With the deepening of information construction and the advancement of professional division of labor, enterprises have shifted from simply focusing on digital construction to focusing on the data content itself. The importance and value of data have gradually emerged, especially in the fields of strategic planning and risk management. With the help of in-depth analysis and mining of data, the decision-making and analysis capabilities of enterprises can be effectively improved. Enterprises continue to pay attention to the standardization of high-tech assets and the quality of assets. The development of high-tech asset management has gradually changed from scattered and independent work to coordinated and integrated work. Enterprises have also established management departments to lead the technical data governance work. However, high-tech asset management still faces many problems: (1) Data dispersion and information fragmentation. The basic data, training data, model data, application data, etc. of current technical assets are scattered in different application systems, and the data exist in different formats such as structured forms, offline Excel, and online PDF. The entire data chain around technology cannot be established, which affects the comprehensive analysis of assets and affects the further improvement of integrity management. (2) Duplicate storage and poor quality. All businesses have different degrees of repeated entry and duplicate storage of technical data. In different systems, the fields and formats of basic technical information are prone to data inconsistency due to different input methods and business needs. The input workload is large, and the efficiency of data sharing and utilization is low. (3) Fragmented management and lack of system. The current high-tech asset integrity management lacks coordinated management of technical data, and has not yet formed a complete management directory, which can easily lead to a situation where it is difficult to accurately query. (4) Technical data security faces challenges. The data security issues caused by the immaturity of asset management technology are also very prominent, and there is a risk of technical data leakage and abuse. Summary of the invention

[0004] Purpose of the invention: In view of the above shortcomings, the present invention provides a high-tech data asset management method based on multi-feature fusion to improve management efficiency.

[0005] Technical solution: To solve the above problems, the present invention adopts a high-tech data asset management method based on multi-feature fusion, which includes the following steps:

[0006] (1) Build a management framework for high-tech data assets based on the actual needs of high-tech data asset management;

[0007] (2) acquiring data related to high-tech data assets and preprocessing the data, wherein the data includes image data and text data;

[0008] (3) extracting local features and global features from the preprocessed image data, and fusing the local features and global features to obtain an image feature vector; extracting features from the preprocessed text data to obtain a text feature vector;

[0009] (4) The text feature vector and the image feature vector are fused through a cross-modal attention mechanism to obtain a text-image feature fusion vector after local feature fusion, and a multi-head attention mechanism is used to obtain a multi-modal fusion global feature;

[0010] (5) Classify data according to the global characteristics of multimodal fusion to form an asset catalog of high-tech data assets.

[0011] Furthermore, the preprocessing of the image data includes cropping, random rotation, flipping, scaling, random cropping of sub-regions, color conversion, noise removal, and normalization; the preprocessing of the text data includes text cleaning, removal of HTML tags and special symbols, unified format, removal of abnormal text, segmentation into words or terms, extraction of stems, and restoration of word forms.

[0012] Furthermore, the image feature vector calculation formula is:

[0013]

[0014] Emb IMG = gelu(W IMG E IMG +b IMG )

[0015] in, The pre-trained weights for the model, represents the local feature vector of the image, Represents the global feature vector of the image, E IMGis the feature vector obtained by concatenating the local feature vector of the image and the global feature vector of the image, Emb IMG is the final feature vector of the image data, W IMG is the training weight matrix, b IMG is the bias term, gelu is the activation function, and I is the input image.

[0016] Furthermore, the text feature vector calculation formula is:

[0017]

[0018] Emb TEXT = gelu(W TEXT E TEXT +b TEXT )

[0019] in, Pre-trained weights for the model, Emb TEXT is the final feature vector of the text data, W TEXT is the training weight matrix, b TEXT is the bias term, T is the input text, ELECTRA is the feature extraction model, E TEXT It is the text feature vector extracted by the feature extraction model.

[0020] Furthermore, the specific steps of fusing the text feature vector and the image feature vector through the cross-modal attention mechanism to obtain the image-text feature fusion vector after local feature fusion are as follows:

[0021] Calculate the text feature vector Emb separately TEXT and image feature vector Emb IMG The cross-modal attention score matrix is ​​used to calculate the text feature vector Emb respectively. TEXT For the image feature vector Emb IMG The weight matrix and image feature vector Emb IMG For text feature vector Emb TEXT According to the weight matrix of text features and image features, the text feature vector matrix Emb guided by image features is obtained respectively. TEXT ' and the image feature vector matrix Emb guided by text features IMG '; Emb TEXT '、Emb IMG '、Emb TEXT and Emb IMG After splicing, the Transformar encoder is input for feature fusion, and finally the image and text feature fusion vector Emb after local feature fusion is obtained c .

[0022] Furthermore, the multi-head attention mechanism is used to obtain the improved text feature vector Emb in the multi-modal fusion global feature by processing the text based on the multi-head attention mechanism. mTEXT :

[0023]

[0024] Among them, h (E) represents the embedding vector of the hidden layer of the ELECTRA model, and represents the weight matrix, and represents the bias vector;

[0025] Embed mTEXT 、Emb TEXT '、Emb IMG ' and Emb IMG After splicing, the feature fusion is input into the Transformer encoder to obtain the image and text feature fusion vector Emb after global feature fusion c '.

[0026] Furthermore, the calculation formula of the multi-attention mechanism is as follows:

[0027]

[0028] in, represents the i-th attention head, Q E , K E 、V E denote the query, key, and value matrices of the embedding vector of the ELECTRA feature extraction model, respectively, and W i Q , W i K , W i V denote the weights of the query, key, and value of the i-th attention head respectively; denote the query, key, and value matrices after the linear transformation of the i-th attention head, respectively, and d k represents the dimension of the key matrix; h represents the number of queries, W O For additional weight.

[0029] Furthermore, a security access strategy is adopted for the asset catalog of the high-tech data assets formed, and the security access strategy includes:

[0030] Trusted identity authentication: Establish a unique and trusted digital identity for asset management users, which includes the legal certificate information of the ID card; combine passwords, dynamic tokens, and biometrics for multi-factor authentication;

[0031] Data encryption transmission: End-to-end encryption is used to ensure that technical data is always encrypted during transmission, and there is no decryption process from sender to receiver;

[0032] Smart contracts: Use smart contracts to control access rights to data assets, so that only authorized users can access data assets;

[0033] Access control: Assign the minimum access rights to each user to ensure that the user can only access the data asset directory and corresponding data required for his / her work; at the same time, record the user's access to data assets; use machine learning technology to detect abnormal behavior of users accessing data assets, including abnormal access frequency and access time, and issue warnings and handle them in a timely manner;

[0034] Security operation and maintenance: Conduct security audits regularly to check the security of data assets, and promptly discover and repair security vulnerabilities; and formulate emergency plans.

[0035] The present invention also adopts a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the above method when executing the computer program.

[0036] The present invention also adopts a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps of the above method are implemented.

[0037] Beneficial effect: Compared with the prior art, the significant advantage of the present invention is that it establishes a multimodal feature fusion model for high-tech assets. Compared with the traditional feature extraction model, it realizes cross-modal image and text feature fusion extraction, mines the respective information, common information and related information of images and texts, uses the multi-attention mechanism to improve text features, and better fuses image and text features, so that the feature extraction model has higher accuracy, better helps enterprises to establish technology asset catalogs, and improve asset management efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 Shown is a flow chart of the high-tech data asset management method of the present invention. DETAILED DESCRIPTION

[0039] like Figure 1 As shown, a high-tech data asset management method based on multi-feature fusion in this embodiment includes the following steps:

[0040] Step 1: Build a management framework for high-tech data assets based on the actual needs of high-tech data asset management;

[0041] Step 2: Acquire and pre-process data related to high-tech data assets, including image data and text data;

[0042] Use RESTful API, SOAP and other technologies to collect heterogeneous technical data from different knowledge base systems, project management systems, and scientific research management systems, and divide the data into image and text categories for standardized processing.

[0043] Step 2-1: Image data integration and preprocessing.

[0044] First, read and decode the image. Use OpenCV, Pillow and other libraries to read and display image data in various formats such as JPEG, PNG, BMP, etc. Then, use the cv2.resize() function to resize the image to fit the model input and crop specific areas of the image as needed. Thirdly, enhance the image by randomly rotating, flipping, scaling, randomly cropping sub-regions, and color transformation, and perform noise processing and outlier monitoring on the image to remove image noise, detect and process damaged files, and extreme exposure. Finally, normalize the image. Scaling the image pixel values ​​to the [0,1] or [-1,1] interval helps the convergence of model training.

[0045] Step 2-2: Text data integration and preprocessing.

[0046] First, read text data in various formats such as CSV, JSON, XML, PDF, WORD, CAJ, etc. Then, clean the text, including removing HTML tags, special symbols and other useless characters, converting the text to a unified lowercase format, detecting and processing abnormal text such as repeated text, short text, non-text, etc. Thirdly, split the text into words or phrases, remove stop words, and use Porter Stemm e And Lemmatization extracts stems and restores word forms. Finally, the text is converted into vector form.

[0047] Step 3: Feature extraction based on multimodal feature fusion technology.

[0048] The multimodal feature fusion technology is used to extract effective features from high-tech data. First, the local and global features of the image are extracted using the ResNet model, and the improved text data features are extracted using the ELECTRA model, which are then fused through a cross-modal attention mechanism. Then, the local features are further fused using the Transformer encoder. Finally, the multi-head attention mechanism is used to obtain the multimodal fused global features.

[0049] Step 3-1: High-tech multimodal data feature extraction.

[0050] (1) For image data, the pre-trained ResNet model is used to extract local and global features of the image. After the local and global features are concatenated, the Gel u The activation function obtains the final image feature vector Emb IMG .

[0051]

[0052] Emb IMG = gelu(W IMG E IMG +b IMG ) (2)

[0053] in, The pre-trained weights for the model, represents the local feature vector of the image, Represents the global feature vector of the image, Emb IMG is the final feature vector of the image data, W IMG is the training weight matrix, b IMG is the deviation term.

[0054] (2) For text data, use the pre-trained ELECTRA model to extract the feature vector Emb of the text TEXT Compared with the BERT model, the ELECTRA model can solve the problem of information mismatch between pre-training tasks and downstream tasks. TEXT The final text feature vector Emb is obtained through linear transformation and Gelu activation function TEXT .

[0055]

[0056] Emb TEXT = gelu(W TEXT E TEXT +b TEXT ) (4)

[0057] in, Pre-trained weights for the model, Emb TEXT is the final feature vector of the text data, W TEXT is the training weight matrix, b TEXT is the deviation term.

[0058] Step 3-2: Fusion of local features of high-tech data.

[0059] (1) Cross-modal attention mechanism matrix

[0060] Calculate the text feature vector Emb separately TEXT and image feature vector EmbIMG The cross-modal attention score matrix is ​​used to calculate the text feature vector Emb respectively according to the cross-modal attention score matrix. TEXT For the image feature vector Emb IMG The weight matrix and image feature vector Emb IMG For text feature vector Emb TEXT The weight matrix of the text feature is obtained by the inner product of the text feature vector and the text feature matrix Emb. TEXT ', the product of the image feature weight matrix and the image feature vector is the image feature vector matrix Emb IMG '.

[0061] (2) Local feature fusion

[0062] Embed TEXT '、Emb IMG '、Emb TEXT and Emb IMG After splicing, it is input into the Transformar encoder for feature fusion.

[0063] Transformar encoder dynamically weights Emb through self-attention mechanism TEXT '、Emb IMG '、Emb TEXT and Emb IMG , further discover the connection and common information between images and texts, as well as the unique information of images and texts, so as to obtain better feature fusion effect. Finally, the image and text feature fusion vector Emb after local feature fusion is obtained C .

[0064] Emb C =Trans(Concat(Emb TEXT ,Emb TEXT ',Emb IMG ,Emb IMG ')) (5)

[0065] Among them, Emb C It represents the vector obtained by fusing the local features of the image and text, and Trans represents the Transformer network.

[0066] Step 3-3: Global feature fusion of high-tech data.

[0067] 91) Text processing based on multi-head attention mechanism

[0068] The multi-attention mechanism is used to better explain the semantic relationship of each word context, thereby improving the model's ability to capture long-distance dependencies and semantic information. The calculation formula of the multi-attention mechanism is as follows:

[0069]

[0070] in, represents the i-th attention head, Q E , K E , V E denote the query, key, and value matrices of the ELECTRA embedding vector, respectively, and W i Q , W i K , W i V denote the weights of the query, key, and value of the i-th attention head respectively; denote the query, key, and value matrices after the linear transformation of the i-th attention head, respectively, and d k represents the dimension of the key matrix; h represents the number of queries, W O For additional weight.

[0071] In order to solve the problems of gradient loss and gradient explosion, residual connection, layer normalization and forward neural network are used for nonlinear transformation and feature extraction. u Activation function, and finally calculate the improved text feature vector Emb mTEXT The calculation formula is as follows:

[0072]

[0073] Among them, h (E) represents the embedding vector of the hidden layer of the ELECTRA model, and They represent the weight matrices, and They represent the bias vectors respectively.

[0074] (2) Global feature fusion

[0075] Embed mTEXT 、Emb TEXT '、Emb IMG ' and Emb IMG After splicing, it is input into the Transformer encoder for feature fusion.

[0076] The Transformer encoder dynamically weights the Embed through the self-attention mechanism. mTEXT 、Emb TEXT '、Emb IMG ' and Emb IMG , further discover the connection and common information between images and texts, as well as the unique information of images and texts, so as to obtain better feature fusion effect. Finally, the image and text feature fusion vector Emb after global feature fusion is obtainedC '.

[0077] Emb C '=T(Concat(Emb mTEXT ,Emb TEXT ',Emb IMG ,Emb IMa ')) (10)

[0078] Step 4: Generate an asset catalog based on high-tech data features.

[0079] First, the extracted high-tech image-text fusion features are classified according to certain rules, such as type, source, application scenario, etc., to facilitate subsequent management and use.

[0080] Secondly, define the feature attributes and define a series of attributes for each feature. Organize the feature attribute information into a data dictionary as the basis for the asset catalog. Feature attributes include but are not limited to feature name, feature description, feature type, data type, data format, source system, update frequency, quality assessment, and data security category. As shown in the following table:

[0081] Table | Data feature attribute statistics example

[0082] Feature attributes illustrate Feature Name Used to uniquely identify features Feature Description Describe the meaning, origin, purpose, etc. of the feature in detail Feature Type For example, numeric type, category type, text type, etc. Data Types For example, integers, floating point numbers, strings, etc. Data Format For example, CSV, JSON, XML, etc. Source System The system or platform where the feature data comes from Update frequency Update frequency of feature data Quality Assessment Assess the quality of feature data Data security level Security level of characteristic data

[0083] Finally, generate a high-tech asset catalog. Design the structure of the asset catalog based on data feature classification and business needs, and organize it by high-tech themes, business lines, data types, etc. Define a series of metadata for the catalog, including but not limited to the catalog name, catalog description, catalog type, catalog version, and catalog owner used to uniquely identify the catalog. Create an index for the catalog based on the catalog metadata to facilitate users to quickly find the required technical data features.

[0084] Step 5: High-tech data asset security access strategy.

[0085] In order to ensure the security of high-tech data assets, a comprehensive security access strategy based on trusted identity authentication, data encryption transmission, smart contracts, access control, and secure operation and maintenance is adopted. By using trusted identity identification, multi-factor authentication, end-to-end encryption, smart contracts, the principle of least privilege, access auditing, abnormal behavior detection, security auditing and other measures, a multi-level, comprehensive security protection system is built to effectively resist various security threats and ensure the security and controllability of data assets. The entire security access strategy process is:

[0086] (1) Trusted identity authentication: Establish a unique and trusted digital identity for asset management users to ensure that the identity of the subject accessing data assets is authentic and reliable. The identity includes legal ID information, and is authenticated in cooperation with a nationally designated identity authentication agency, and supports traceability.

[0087] In addition, authentication is performed by combining passwords, dynamic tokens, and biometrics to improve the security of identity authentication.

[0088] (2) Data encryption transmission: End-to-end encryption is used to ensure that technical data is always encrypted during transmission, with no decryption process from sender to receiver, thus avoiding the leakage of confidential technical data.

[0089] (3) Smart contracts: Use smart contracts to control access rights to data assets, ensuring that only authorized users can access data assets. Smart contracts can automatically enforce data access rules and reduce human errors.

[0090] (4) Access control: Assign the minimum access rights to each user to ensure that the user can only access the data asset directory and corresponding data required for his / her work. At the same time, record the user's access to data assets for security audit and traceability. Use machine learning technology to detect abnormal behavior of users accessing data assets, including abnormal access frequency and access time, and issue warnings and handle them in a timely manner.

[0091] (5) Security Operation and Maintenance: Conduct security audits regularly to check the security of data assets and promptly discover and repair security vulnerabilities. Develop emergency plans so that security incidents can be handled in a timely manner.

[0092] The present invention constructs a multimodal feature fusion model of high-tech assets based on ResNet and multi-attention mechanism, realizes the automatic extraction and label definition of multi-feature fusion of multi-source heterogeneous technology assets, realizes intelligent combing of technology assets, reduces manual operations, and improves data processing efficiency.

[0093] Using image processing algorithms and natural language processing algorithms, we integrate and process intellectual property-related technical data from different sources and formats, break data silos, achieve unified management and sharing of high-tech assets, and improve the efficiency of high-tech data governance. Based on the multimodal feature fusion model, we automatically generate a catalog of enterprise high-tech assets to facilitate enterprise management and utilization of technology assets. We propose a secure access strategy for high-tech asset management, which covers the entire asset management life cycle of trusted identity authentication, data encryption transmission, smart contracts, access control, and secure operation and maintenance to ensure the security and controllability of technical data.

[0094] The present invention adopts data mining technology to construct a management framework based on the fusion of multi-source heterogeneous high-tech data features, integrates and processes intellectual property-related technical data, and uses a multimodal feature fusion model to intelligently extract technology asset data features and define data labels, generate an enterprise high-tech asset catalog, and design an asset management security access strategy, clearly sort out the knowledge structure of technology assets, improve the efficiency of enterprise technology asset management, and assist the formulation and decision-making of the enterprise's core technology development direction and strategy.

[0095] Based on the physical characteristics of multi-source heterogeneous high-tech assets, a high-tech asset integration and standardization processing mechanism is established; based on the high-tech asset integration and standardization processing mechanism, standardized heterogeneous data is further obtained; using standardized heterogeneous data, a multimodal feature fusion model based on ResNet and multi-attention mechanism is proposed to achieve high-tech asset feature extraction and label definition, thereby generating a high-tech asset catalog based on asset features. The invention supplements the existing high-tech asset management methods and provides an effective means for forming core technology analysis reports and laying out patent strategies in the field of high-tech technology.

Claims

1. A high-tech data asset management method based on multi-feature fusion, characterized in that: The following steps are involved: (1) Build a management framework for high-tech data assets based on the actual needs of high-tech data asset management; (2) acquiring data related to high-tech data assets and preprocessing the data, wherein the data includes image data and text data; (3) extracting local features and global features from the preprocessed image data, and fusing the local features and global features to obtain an image feature vector; extracting features from the preprocessed text data to obtain a text feature vector; (4) The text feature vector and the image feature vector are fused through a cross-modal attention mechanism to obtain a text-image feature fusion vector after local feature fusion, and a multi-head attention mechanism is used to obtain a multi-modal fusion global feature; (5) Classify data according to the global characteristics of multimodal fusion to form an asset catalog of high-tech data assets.

2. The high-tech data asset management method according to claim 1, characterized in that: The preprocessing of the image data includes cropping, random rotation, flipping, scaling, random cropping of sub-regions, color conversion, noise removal, and normalization; the preprocessing of the text data includes text cleaning, removal of HTML tags and special symbols, unified format, removal of abnormal text, segmentation into words or terms, extraction of stems, and restoration of word forms.

3. The high-tech data asset management method according to claim 2, characterized in that: The image feature vector calculation formula is: Emb IMG Jgelu(W IMG E IMG +b IMG ) in, The pre-trained weights for the model, represents the local feature vector of the image, Represents the global feature vector of the image, E IMG is the feature vector obtained by concatenating the local feature vector of the image and the global feature vector of the image, Emb IMG is the final feature vector of the image data, W IMG is the training weight matrix, b IMG is the bias term, gelu is the activation function, and I is the input image.

4. The high-tech data asset management method according to claim 3 is characterized in that: The text feature vector calculation formula is: Emb TEXT Jgelu(W TEXT E TEXT +b TEXT ) in, Pre-trained weights for the model, Emb TEXT is the final feature vector of the text data, W TEXT is the training weight matrix, b TEXT is the bias term, T is the input text, ELECTRA is the feature extraction model, E TEXT It is the text feature vector extracted by the feature extraction model.

5. The high-tech data asset management method according to claim 4, characterized in that: The specific steps of fusing the text feature vector and the image feature vector through the cross-modal attention mechanism to obtain the image-text feature fusion vector after local feature fusion are as follows: Calculate the text feature vector Emb separately TEXT and image feature vector Emb IMG The cross-modal attention score matrix is ​​used to calculate the text feature vector Emb respectively. TEXT For the image feature vector Emb IMG The weight matrix and image feature vector Emb IMG For text feature vector Emb TEXT According to the weight matrix of text features and image features, the text feature vector matrix Emb guided by image features is obtained respectively. TEXT ' and the image feature vector matrix Emb guided by text features IMG '; Embed the text feature vector matrix TEXT ', image feature vector matrix Emb IMG ', text feature vector Emb TEXT and image feature vector Emb IMG After splicing, the Transformar encoder is input for feature fusion, and finally the image and text feature fusion vector Emb after local feature fusion is obtained C .

6. The high-tech data asset management method according to claim 5, characterized in that: The multi-head attention mechanism is used to obtain the multi-modal fusion global feature, and the improved text feature vector Emb is obtained by processing the text based on the multi-head attention mechanism. mTEXT : Among them, h (E) represents the embedding vector of the hidden layer of the ELECTRA model, and represents the weight matrix, and represents the bias vector; The improved text feature vector Emb mTEXT , text feature vector matrix Emb TEXT ', Image feature vector matrix Emb IMG ' and image feature vector Emb IMG After splicing, the feature fusion is input into the Transformer encoder to obtain the image and text feature fusion vector Emb after global feature fusion C '.

7. The high-tech data asset management method according to claim 6, characterized in that: The calculation formula of the multi-attention mechanism is as follows: in, represents the i-th attention head, QQ E , K E 、V E denote the query, key, and value matrices of the embedding vector of the ELECTRA feature extraction model, respectively, and W i Q , W i K , W i V denote the weights of the query, key, and value of the i-th attention head respectively; V i E denote the query, key, and value matrices after the linear transformation of the i-th attention head, respectively, and d k represents the dimension of the key matrix; h represents the number of queries, W O For additional weight.

8. The high-tech data asset management method according to claim 7, characterized in that: A security access strategy is adopted for the asset catalog of the high-tech data assets formed, and the security access strategy includes: Trusted identity authentication: Establish a unique and trusted digital identity for asset management users, which includes the legal certificate information of the ID card; combine passwords, dynamic tokens, and biometrics for multi-factor authentication; Data encryption transmission: End-to-end encryption is used to ensure that technical data is always encrypted during transmission, and there is no decryption process from sender to receiver; Smart contracts: Use smart contracts to control access rights to data assets, so that only authorized users can access data assets; Access control: Assign the minimum access rights to each user to ensure that the user can only access the data asset directory and corresponding data required for his / her work; at the same time, record the user's access to data assets; use machine learning technology to detect abnormal behavior of users accessing data assets, including abnormal access frequency and access time, and issue warnings and handle them in a timely manner; Security operation and maintenance: Conduct security audits regularly to check the security of data assets, and promptly discover and repair security vulnerabilities; and formulate emergency plans.

9. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 8 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 8 are implemented.

Images