A method and device for preventing data leakage in rail transit based on matrix transformation

Abstract

This invention provides a method and apparatus for preventing data leakage in rail transit based on matrix transformation, relating to the field of rail transit data security technology. The method includes converting rail transit data into matrix data; each row of the matrix data represents a data record, and each column represents a data attribute; generating an invertible random matrix based on the number of columns in the matrix data; matching the number of rows and columns of the random matrix with the number of columns in the matrix data; multiplying the matrix data with the random matrix to obtain obfuscated matrix data; evaluating the privacy protection strength of the obfuscated matrix data until the privacy protection strength meets a preset standard; and when a usage instruction for the rail transit data is received, determining the inverse matrix of the random matrix and multiplying the inverse matrix with the obfuscated matrix data to obtain the restored rail transit data. This significantly reduces the risk of data leakage during storage and transmission, ensures the integrity of the data during legitimate use, and achieves a balance between privacy protection and data availability.

Description

Technical Field

[0001] This invention relates to the field of rail transit data security technology, and in particular to a method and apparatus for preventing the leakage of rail transit data based on matrix transformation. Background Technology

[0002] With the rapid development of the rail transit industry, a large amount of rail transit data is generated and transmitted during system operation, maintenance, and management. This data often contains sensitive information, such as passenger personal information, train operating status, and dispatch instructions. If this data is leaked, it could lead to serious consequences such as passenger privacy breaches, increased operational safety risks, and damage to the company's reputation.

[0003] Existing data protection methods mainly focus on data encryption and access control. However, these methods may have the following problems when facing the complex rail transit data environment. For example, Problem 1: Although traditional encryption methods can encrypt data, decryption is required during data use, which may lead to the risk of data leakage during transmission and processing after decryption; Problem 2: Relying solely on access control is insufficient to completely prevent data leakage caused by internal personnel or system vulnerabilities; Problem 3: Rail transit data has a combination of structured and unstructured characteristics, and traditional protection methods are difficult to process different types of data uniformly. Summary of the Invention

[0004] This invention provides a method and apparatus for preventing leakage of rail transit data based on matrix transformation, which solves the technical problem that existing technologies cannot effectively protect rail transit data.

[0005] On the one hand, this invention provides a method for preventing data leakage in rail transit based on matrix transformation, including:

[0006] The rail transit data is converted into matrix data; wherein each row of the matrix data represents a data record and each column represents a data attribute.

[0007] Based on the number of columns in the matrix data, an invertible random matrix is ​​generated; wherein the number of rows and columns of the random matrix matches the number of columns in the matrix data.

[0008] Multiply the matrix data by the random matrix to obtain the confusion matrix data;

[0009] Evaluate the privacy protection strength of the obfuscation matrix data until the privacy protection strength meets a preset standard;

[0010] When a command to use rail transit data is received, the inverse of the random matrix is ​​determined, and the inverse matrix is ​​multiplied by the confusion matrix data to obtain the restored rail transit data.

[0011] According to the present invention, a method for preventing data leakage in rail transit based on matrix transformation converts rail transit data into matrix data, including:

[0012] Determine the data type of each data attribute of the rail transit data, and use the mapping rule corresponding to the data type to map each data attribute to a column of matrix data;

[0013] Arrange each data record according to its relationship to form a row of matrix data;

[0014] The step of determining the data type of each data attribute of the rail transit data, and mapping each data attribute to a column of matrix data using a mapping rule corresponding to the data type, includes:

[0015] When the data attribute is numerical, its value is directly used as a matrix element;

[0016] When the data attributes are categorical, one-hot encoding is used to convert them into numerical data, which are then used as matrix elements.

[0017] The correlation includes temporal order or spatial order.

[0018] According to the present invention, a method for preventing data leakage in rail transit based on matrix transformation is provided, which generates an invertible random matrix according to the number of columns of the matrix data, including:

[0019] Determine the dimensions of the random matrix so that its number of rows and columns matches the number of columns in the matrix data;

[0020] Determine the distribution characteristics of rail transit data, and use a random generation algorithm corresponding to the distribution characteristics to generate elements of a random matrix;

[0021] Check if a random matrix is ​​invertible by calculating its determinant or rank. If it is not invertible, regenerate it until an invertible random matrix is ​​obtained.

[0022] The determination of the distribution characteristics of rail transit data, including generating elements of a random matrix using a random generation algorithm corresponding to the distribution characteristics, comprises:

[0023] When the distribution characteristics of rail transit data are normally distributed, the element values ​​of the random matrix are generated using the normal distribution method.

[0024] When the distribution characteristics of rail transit data are uniform, the element values ​​of the random matrix are generated using the uniform distribution method.

[0025] According to the present invention, a method for preventing data leakage in rail transit based on matrix transformation is provided, which multiplies the matrix data with the random matrix to obtain obfuscated matrix data, including:

[0026] The matrix size is determined based on the matrix data and the number of rows and columns of the random matrix;

[0027] When the size of the matrix is ​​less than the minimum value of the preset size range, the matrix data is directly multiplied by the random matrix;

[0028] When the size of the matrix falls within the preset size range, the matrix data is multiplied by the random matrix using either the Strassen algorithm or the Winograd algorithm.

[0029] When the size of the matrix is ​​greater than the maximum value of the preset size range, the matrix data and the random matrix are divided into multiple sub-matrix blocks respectively;

[0030] Perform matrix multiplication on each sub-matrix block to obtain the corresponding sub-result block;

[0031] All sub-result blocks are concatenated according to their positions in the original matrix to form a complete confusion matrix.

[0032] According to a matrix transformation-based method for preventing data leakage in rail transit provided by the present invention, the privacy protection strength of the obfuscated matrix data is evaluated until the privacy protection strength meets a preset standard, including:

[0033] The assessment indicators for the strength of privacy protection are determined based on the characteristics of rail transit data and the needs for privacy protection.

[0034] Calculate the privacy protection strength value of the confusion matrix data based on the evaluation indicators;

[0035] Compare the privacy protection strength value with a preset standard;

[0036] If the privacy protection strength value does not meet the preset standard, a random matrix is ​​regenerated, increasing the complexity of the random matrix, or noise is added to the confused matrix data to enhance the privacy protection strength value until the privacy protection strength meets the preset standard.

[0037] According to the present invention, a method for preventing data leakage in rail transit based on matrix transformation is provided. This method determines evaluation indicators for the strength of privacy protection based on the characteristics of rail transit data and privacy protection requirements, including:

[0038] Assess the sensitivity of each data attribute in rail transit data, and determine the corresponding assessment index weights based on the level of sensitivity;

[0039] For rail transit data where the proportion of categorical or numerical data exceeds a preset ratio, information entropy is selected as the evaluation index. The strength of privacy protection is quantified by calculating the probability distribution of each element value in the confusion matrix data. The higher the information entropy, the stronger the privacy protection.

[0040] When it is necessary to assess the correlation between confusion matrix data and matrix data, mutual information is selected as the assessment metric; the lower the mutual information, the higher the strength of privacy protection.

[0041] For rail transit data that needs to meet differential privacy protection requirements, the privacy budget value of differential privacy is selected as the evaluation index; the smaller the privacy budget value, the higher the privacy protection strength.

[0042] According to the present invention, a method for preventing data leakage in rail transit based on matrix transformation, after evaluating the privacy protection strength of the obfuscated matrix data until the privacy protection strength meets a preset standard, further includes:

[0043] The confusion matrix data is fragmented according to preset rules to generate multiple data segments;

[0044] Each data segment is stored on a different storage node or storage medium to ensure isolation between data segments;

[0045] When receiving instructions to use rail transit data, the following is also included:

[0046] Collect all data fragments and perform restoration processing.

[0047] According to the present invention, a method for preventing leakage of rail transit data based on matrix transformation is provided, which further includes, before converting the rail transit data into matrix data:

[0048] Distinguish between structured and unstructured data in rail transit data;

[0049] Combine the structured data with the unstructured data;

[0050] The manner of combination includes at least one of the following:

[0051] Feature concatenation, weighted averaging, or feature extraction and fusion from deep learning models.

[0052] According to the present invention, a method for preventing data leakage in rail transit based on matrix transformation, after generating an invertible random matrix, further includes:

[0053] A quantum entangled state generator is used to inject quantum entangled state information into each element of the random matrix; wherein the quantum entangled state information is represented by qubits, and the state of each qubit is determined by the element value of the random matrix;

[0054] Before multiplying the matrix data with the random matrix, the method further includes:

[0055] Quantum state verification is performed on the random matrix injected with quantum entangled states to ensure that the quantum state information of each element is complete and has not been tampered with.

[0056] On the other hand, the present invention also provides a data leakage prevention processing device for rail transit based on matrix transformation, comprising:

[0057] The matrix conversion module is used to convert rail transit data into matrix data; wherein each row of the matrix data represents a data record and each column represents a data attribute.

[0058] A random matrix module is used to generate an invertible random matrix based on the number of columns in the matrix data; wherein the number of rows and columns of the random matrix matches the number of columns in the matrix data;

[0059] A matrix mixing module is used to multiply the matrix data with the random matrix to obtain confused matrix data;

[0060] The matrix evaluation module is used to evaluate the privacy protection strength of the obfuscated matrix data until the privacy protection strength meets a preset standard.

[0061] The matrix restoration module is used to determine the inverse matrix of the random matrix when a usage instruction for rail transit data is received, and multiply the inverse matrix with the confusion matrix data to obtain the restored rail transit data.

[0062] The present invention provides a method and apparatus for preventing data leakage in rail transit based on matrix transformation. By converting rail transit data into matrix form and using an invertible random matrix for obfuscation, the characteristics of the original data can be effectively hidden, significantly reducing the risk of data leakage during storage and transmission. Simultaneously, data restoration is performed using the inverse of the random matrix, ensuring the integrity and availability of the data during legitimate use. This achieves a balance between privacy protection and data availability, providing an efficient and reliable method for the secure processing of rail transit data. Attached Figure Description

[0063] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0064] Figure 1 This is a flowchart illustrating the data leakage prevention method for rail transit based on matrix transformation provided in an embodiment of the present invention.

[0065] Figure 2 This is a schematic diagram of the structure of the rail transit data leakage prevention and control device based on matrix transformation provided in an embodiment of the present invention;

[0066] Figure 3 This is a schematic diagram of the structure of the electronic device provided in an embodiment of the present invention. Detailed Implementation

[0067] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0068] Figure 1 This is a flowchart illustrating the data leakage prevention method for rail transit based on matrix transformation provided in this embodiment of the invention. The method can be executed by a computer, tablet computer, or smart wearable device, etc.

[0069] See Figure 1 The method for preventing data leakage in rail transit based on matrix transformation may include the following steps 101 to 105.

[0070] Step 101: Convert the rail transit data into matrix data; where each row of the matrix data represents a data record and each column represents a data attribute.

[0071] In this step, the rail transit data may include passenger personal information, train operation status data, dispatching instruction data, ticketing system data, equipment operation data, and operation management data. Passenger personal information may include name, ID number, contact information, ticketing information, and payment information. Train operation status data may include train location information, train speed, train status, carriage status, and train timetable. Dispatching instruction data may include dispatching orders and dispatching plans. Ticketing system data may include ticket sales data, ticket inspection data, and discount information. Equipment operation data may include track equipment status, signaling system data, vehicle equipment status, and station equipment status. Operation management data may include passenger flow data, operating revenue data, operating cost data, and security check data.

[0072] For example, a data record may contain Zhang San's ticketing information, and the data attributes may include name (Zhang San), ID number, train number, seat number, etc.; a data record may contain station passenger flow statistics, and the data attributes may include the number of people entering the station, the number of people exiting the station, the statistical time, etc.

[0073] Step 101 may specifically include:

[0074] Determine the data type of each data attribute of the rail transit data, and use the mapping rule corresponding to the data type to map each data attribute to a column of matrix data;

[0075] Arrange each data record according to its relationship to form a row of matrix data;

[0076] This involves determining the data type of each data attribute in the rail transit data, and using a mapping rule corresponding to the data type to map each data attribute to a column of matrix data, including:

[0077] When the data attribute is numerical, its value is directly used as a matrix element;

[0078] When the data attributes are categorical, one-hot encoding is used to convert them into numerical data, which are then used as matrix elements.

[0079] Relationships can be chronological or spatial.

[0080] Time sequence: Data records are arranged in chronological order. For example, train operation status data can be arranged in order of train operation time; passenger ticketing information can be arranged in order of ticket purchase time.

[0081] Spatial order: Data records are arranged according to their spatial location. For example, passenger flow statistics for a station can be arranged according to the station's geographical location; seat information for a train can be arranged according to the physical location of the seats.

[0082] This step clarifies the specific steps for converting rail transit data into matrix data, including the processing methods for different data types (direct conversion of numerical data and one-hot encoding for categorical data) and the arrangement of data records (by time or spatial order). This processing method not only unifies complex data structures into matrix form, facilitating subsequent matrix transformation operations, but also preserves the original characteristics and correlations of the data, providing a solid foundation for subsequent privacy protection processing. It enhances the applicability and flexibility of the entire method, enabling it to adapt to diverse data types and application scenarios in the rail transit field.

[0083] Step 102: Generate an invertible random matrix based on the number of columns in the matrix data; wherein the number of rows and columns of the random matrix matches the number of columns in the matrix data.

[0084] Step 102 may specifically include:

[0085] Determine the dimensions of the random matrix so that its number of rows and columns matches the number of columns in the matrix data;

[0086] Determine the distribution characteristics of rail transit data, and use a random generation algorithm corresponding to the distribution characteristics to generate elements of a random matrix;

[0087] The invertibility of a random matrix is ​​checked by calculating its determinant value (e.g., non-zero indicates invertibility) or its rank (e.g., equal to the number of rows or columns). If it is not invertible, it is regenerated until an invertible random matrix is ​​obtained.

[0088] This includes determining the distribution characteristics of rail transit data and generating elements of a random matrix using a random generation algorithm corresponding to these distribution characteristics, including:

[0089] When the distribution characteristics of rail transit data are normally distributed, the element values ​​of the random matrix are generated using the normal distribution method.

[0090] Specifically, it is necessary to determine the mean and standard deviation of the normal distribution, and then use a normal distribution random number generator to generate random numbers.

[0091] The generated random numbers will be used as the element values ​​of the random matrix.

[0092] When the distribution characteristics of rail transit data are uniform, the element values ​​of the random matrix are generated using the uniform distribution method.

[0093] Specifically, it is necessary to determine the range of the uniform distribution interval, and then use a uniformly distributed random number generator to generate random numbers.

[0094] The generated random numbers will be used as the element values ​​of the random matrix.

[0095] This step generates a matching invertible random matrix based on the number of columns in the matrix data. A suitable randomization algorithm (such as a normal or uniform distribution method) is selected based on the distribution characteristics of the rail transit data, enabling the generation of more targeted and adaptive random matrices. The invertibility of the random matrix is ​​checked by calculating the determinant or rank of the matrix, ensuring its reliability and validity and avoiding the problem of data irreversibility due to the non-invertibility of the random matrix. This method not only improves the efficiency of random matrix generation but also enhances data obfuscation, further strengthening and improving the reliability of data privacy protection.

[0096] Step 103: Multiply the matrix data with the random matrix to obtain the confusing matrix data.

[0097] Step 103 may specifically include:

[0098] Determine the matrix size based on the number of rows and columns of the matrix data and the random matrix;

[0099] When the size of the matrix is ​​smaller than the minimum value of the preset size range, the matrix data is directly multiplied by the random matrix;

[0100] When the matrix size falls within a preset size range, the Strassen algorithm or Winograd algorithm is used to multiply the matrix data with a random matrix.

[0101] When the size of the matrix exceeds the maximum value of the preset size range, the matrix data and the random matrix are divided into multiple sub-matrix blocks respectively;

[0102] Perform matrix multiplication on each sub-matrix block to obtain the corresponding sub-result block;

[0103] All sub-result blocks are concatenated according to their positions in the original matrix to form a complete confusion matrix.

[0104] This step addresses matrix multiplication operations of varying scales by employing direct multiplication, optimized algorithms (such as the Strassen or Winograd algorithms), or block processing methods. This allows for flexible selection of the most suitable computational method based on the data size. This not only improves the efficiency of matrix operations and reduces computational resource consumption but also ensures the efficiency and stability of the entire data processing process. This makes the method applicable to privacy-preserving processing of large-scale rail transit data, enhancing its practicality and scalability, and providing strong support for the secure processing of rail transit data in big data environments.

[0105] Step 104: Evaluate the privacy protection strength of the obfuscation matrix data until the privacy protection strength meets the preset standard.

[0106] Step 104 may specifically include:

[0107] The assessment indicators for the strength of privacy protection are determined based on the characteristics of rail transit data and the needs for privacy protection.

[0108] Calculate the privacy protection strength value of the confusion matrix data based on the evaluation indicators;

[0109] Compare the privacy protection strength value with a preset standard;

[0110] If the privacy protection strength value does not meet the preset standard, a random matrix is ​​regenerated, increasing the complexity of the random matrix, or noise is added to the scrambled matrix data to enhance the privacy protection strength value until the privacy protection strength meets the preset standard.

[0111] This step introduces a privacy protection strength assessment mechanism. By setting assessment indicators and preset standards, it quantifies the privacy protection effect of the obfuscated matrix data. This method can monitor the data privacy protection level in real time, ensuring that it meets the privacy protection needs of practical applications. When the privacy protection strength fails to meet the preset standards, adjustments can be made by regenerating the random matrix, increasing the complexity of the random matrix, or adding noise to the obfuscated matrix data. This dynamically optimizes privacy protection measures, further enhancing the data privacy protection effect and improving the reliability and adaptability of the entire method. It allows for flexible adjustments based on different application scenarios and privacy requirements, effectively addressing various data leakage risks.

[0112] Step 105: When the instruction to use rail transit data is received, determine the inverse matrix of the random matrix, and multiply the inverse matrix with the confusion matrix data to obtain the restored rail transit data.

[0113] In this embodiment, by converting rail transit data into matrix form and using an invertible random matrix for obfuscation, the characteristics of the original data can be effectively hidden, significantly reducing the risk of data leakage during storage and transmission. Simultaneously, data restoration using the inverse of the random matrix ensures the integrity and availability of the data during legitimate use, achieving a balance between privacy protection and data availability. This provides an efficient and reliable method for the secure processing of rail transit data.

[0114] In one embodiment of this specification, the evaluation indicators for the strength of privacy protection are determined based on the characteristics of rail transit data and privacy protection requirements, including:

[0115] Step 1: Assess the sensitivity of each data attribute in the rail transit data, and determine the corresponding assessment index weights based on the sensitivity level.

[0116] Assume that the rail transit data includes passengers' personal information (such as ID card numbers and mobile phone numbers) and train operation status data (such as speed and location). ID card numbers and mobile phone numbers are highly sensitive information, while train speed and location, although also important, are relatively less sensitive. Therefore, ID card numbers and mobile phone numbers can be assigned a higher weight (e.g., 0.8), while train speed and location can be assigned a lower weight (e.g., 0.2).

[0117] Step 2: For rail transit data where the proportion of categorical or numerical data exceeds a preset ratio, information entropy is selected as the evaluation index. The strength of privacy protection is quantified by calculating the probability distribution of each element value in the confusion matrix data. The higher the information entropy, the stronger the privacy protection.

[0118] Categorical data refers to data types with a finite number of fixed categories (or labels). Unlike numerical data (such as continuous values ​​like age and income), categorical data typically represents discrete, non-numerical attributes. Even when numerical values ​​are present, they are primarily used to identify categories rather than indicating size or quantity. For example, the range of values ​​for train operation status includes: normal operation, delayed, and cancelled. The range of values ​​for passenger ticket types includes: single-journey ticket, round-trip ticket, monthly pass, and annual pass.

[0119] Step 3: When it is necessary to assess the correlation between the confusion matrix data and the matrix data, mutual information is selected as the assessment indicator; the lower the mutual information, the higher the strength of privacy protection.

[0120] Step 4: For rail transit data that needs to meet differential privacy protection requirements, select the privacy budget value of differential privacy as the evaluation index; where the smaller the privacy budget value, the higher the privacy protection strength.

[0121] This embodiment provides a selection of various privacy protection strength assessment indicators, including information entropy, mutual information, and the privacy budget value of differential privacy. These indicators can quantify the degree of data privacy protection from different perspectives. For example, information entropy is used to measure data uncertainty, mutual information is used to assess the correlation between data, and the privacy budget value of differential privacy is used to meet specific privacy protection requirements. By flexibly selecting assessment indicators according to different data characteristics and privacy needs, the strength of privacy protection can be assessed more accurately, providing a more scientific basis for optimizing privacy protection measures, improving the accuracy and adaptability of the entire method, and enabling it to better meet the privacy protection needs of rail transit data in different scenarios.

[0122] In one embodiment of this specification, after evaluating the privacy protection strength of the confusion matrix data until the privacy protection strength meets a preset standard, the method further includes:

[0123] The confusion matrix data is fragmented according to preset rules to generate multiple data segments;

[0124] Specifically, the sharding rules are determined, such as sharding by row, sharding by column, or sharding by matrix block. For example, a large confusion matrix can be divided into several smaller matrices according to rows, with each smaller matrix containing a subset of rows from the original matrix;

[0125] Based on the partitioning rules, the confusion matrix data is divided into multiple data fragments. For example, assuming the confusion matrix is ​​a 100×100 matrix, it can be divided into 10 smaller 10×100 matrix fragments by row.

[0126] Each data segment is stored on a different storage node or storage medium to ensure isolation between data segments;

[0127] Specifically, multiple storage nodes or storage media can be selected, such as different servers, cloud storage services, or physical storage devices.

[0128] Each data segment is stored separately on a different storage node or storage medium. For example, the first data segment is stored on server A, the second data segment is stored on server B, and so on.

[0129] Ensure the security of each storage node or storage medium, for example, by protecting data fragments through encryption, access control and other measures.

[0130] When receiving instructions to use rail transit data, the following is also included:

[0131] Collect all data fragments and perform restoration processing;

[0132] Specifically, upon receiving a usage instruction, the data restoration process is triggered;

[0133] Collect all data fragments stored on different storage nodes or storage media;

[0134] According to the rules of data fragmentation, all data fragments are spliced ​​together to restore the complete confusion matrix data;

[0135] Determine the inverse of the random matrix, and multiply the inverse matrix with the recovered confusion matrix data to obtain the restored rail transit data.

[0136] In this embodiment, after the privacy protection strength meets the preset standard, the obfuscation matrix data is further fragmented and stored in different storage nodes or storage media, realizing distributed data storage. This storage method increases the isolation of data storage; even if some storage nodes or storage media are compromised, attackers will find it difficult to obtain complete data fragments, thereby further reducing the risk of data leakage. Simultaneously, by collecting all data fragments for restoration processing when data is needed, the integrity and availability of the data are ensured, providing a more reliable guarantee for the secure storage and use of rail transit data and enhancing the overall data security of the method.

[0137] In one embodiment of this specification, before converting rail transit data into matrix data, the method further includes:

[0138] Distinguish between structured and unstructured data in rail transit data;

[0139] Specifically, structured data is typically stored in relational databases and has a clear table structure, such as passenger information (ID number, mobile phone number, ticket purchase time, etc.) and train operation status (train number, speed, location, etc.).

[0140] Unstructured data: It is usually stored in the file system and does not have a fixed table structure, such as surveillance videos, log files, passenger feedback, etc.

[0141] Distinguishing methods: Data can be distinguished by its storage format, data type, and data source. For example, data extracted from a relational database is structured data, while data extracted from a file system is unstructured data.

[0142] Combine structured data with unstructured data;

[0143] The combination method includes at least one of the following:

[0144] Feature concatenation, weighted averaging, or feature extraction and fusion from deep learning models;

[0145] Specifically, feature concatenation involves directly combining features from structured and unstructured data. For example, a passenger's ID number (structured data) and the corresponding surveillance video features (unstructured data) can be concatenated into a single feature vector.

[0146] Weighted average: This involves averaging the features of structured and unstructured data using weighted averages. For example, assigning a weight of 0.7 to structured data and a weight of 0.3 to unstructured data, and then calculating the weighted average.

[0147] Feature extraction and fusion in deep learning models: Deep learning models (such as convolutional neural networks and recurrent neural networks) are used to extract features from structured and unstructured data, and then the extracted features are fused. For example, convolutional neural networks are used to extract features from surveillance videos, and recurrent neural networks are used to extract features from time series data. The two types of features are then fused through a fully connected layer.

[0148] In this embodiment, before converting rail transit data into matrix data, structured and unstructured data are distinguished and combined using methods such as feature concatenation, weighted averaging, or feature extraction and fusion from deep learning models. This processing method fully utilizes the explicit structure of structured data and the rich information of unstructured data, organically integrating different types of data to form a more complete and representative matrix. This not only enriches the data's content and improves its usability but also enhances the adaptability of the entire method to complex data environments, enabling it to better handle the common situation of mixed structured and unstructured data in the rail transit field, thus improving the flexibility and practicality of data processing.

[0149] In one embodiment of this specification, after generating the invertible random matrix, the method further includes:

[0150] A quantum entangled state generator is used to inject quantum entangled state information into each element of a random matrix; where the quantum entangled state information is represented by qubits, and the state of each qubit is determined by the element value of the random matrix;

[0151] Before multiplying the matrix data with the random matrix, the following steps are also included:

[0152] Quantum state verification is performed on the random matrix injected with quantum entangled states to ensure that the quantum state information of each element is complete and has not been tampered with.

[0153] In this embodiment, a quantum entangled state generator is used to inject quantum entangled state information into each element of the random matrix, and quantum state verification ensures the integrity and tamper-proof nature of the information. The introduction of quantum entangled state information adds quantum-level security to the random matrix. Due to the non-cloning and immutability of quantum states, the random matrix exhibits higher security during storage and transmission, further enhancing data obfuscation and privacy protection. Simultaneously, the quantum state verification mechanism effectively detects the integrity of the random matrix during processing, preventing malicious tampering or damage to data during obfuscation. This provides a higher level of security for the privacy protection of rail transit data, improving the overall security and reliability of the method and making it more advantageous in the face of high-level security threats.

[0154] In some other embodiments of this specification, after generating the invertible random matrix, the following steps are also included:

[0155] Quantum encryption is applied to random matrices, utilizing the non-cloning and immutability of quantum states to generate a unique quantum encryption identifier for each element of the random matrix;

[0156] Before multiplying the matrix data with the random matrix, each row of the matrix data is quantum-encoded so that each row of the matrix data corresponds to a quantum state.

[0157] The quantum state-encoded matrix data is multiplied by the quantum encrypted random matrix to obtain quantum confusion matrix data;

[0158] When evaluating the privacy protection strength of confusion matrix data, quantum entanglement is introduced as an evaluation index. The privacy protection strength is quantified by measuring the degree of entanglement between quantum states in the quantum confusion matrix data; the higher the quantum entanglement, the stronger the privacy protection.

[0159] In some other embodiments of this specification, after multiplying the matrix data with a random matrix, the method further includes:

[0160] Blockchain technology is used to perform distributed storage and immutable recording of confusion matrix data. The confusion matrix data is divided into multiple data blocks, and each data block generates a hash value and is stored on various nodes of the blockchain.

[0161] A smart contract is set up on the blockchain. When a command to use rail transit data is received, the smart contract automatically triggers the data restoration process. Only after verifying the integrity of the hash values ​​of each data block on the blockchain is the data restoration operation allowed.

[0162] During the restoration process, the consensus mechanism of blockchain is used to verify the restored rail transit data through multiple nodes to ensure the authenticity and integrity of the data and prevent the data from being tampered with or forged during the restoration process.

[0163] In some other embodiments of this specification, after multiplying the matrix data with the random matrix, the method further includes:

[0164] Step 1: Extract multimodal features from the confusion matrix data, including numerical features, structural features, and time series features, to form a multimodal feature vector;

[0165] Specifically, numerical feature extraction involves calculating the statistical characteristics of each element in the confusion matrix, such as mean, variance, maximum value, and minimum value. These features reflect the distribution of the data.

[0166] Structural feature extraction: Analyzing the structural properties of the confusion matrix, such as its sparsity, symmetry, and row-column correlation. For example, evaluating the linear relationship between data by calculating the row-column correlation coefficient matrix.

[0167] Time series feature extraction: If the confusion matrix data has a time sequence (e.g., rail transit data is recorded by time), then time series features, such as data trends, periodicity, and autocorrelation, can be extracted. Time series analysis methods, such as ARIMA models or wavelet transforms, can be used to extract these features.

[0168] Multimodal feature vector fusion: This combines the three types of features mentioned above into a single multimodal feature vector for subsequent anomaly detection. For example, numerical features, structural features, and time series features can be concatenated into a high-dimensional feature vector with certain weights.

[0169] Step 2: Based on multimodal feature vectors, construct a deep neural network model to dynamically monitor abnormal behavior of confusion matrix data during storage and transmission;

[0170] Specifically, the model architecture design involves designing a deep neural network model, such as using a convolutional neural network (CNN) to process numerical features, a graph neural network (GNN) to process structural features, and a recurrent neural network (RNN) or a long short-term memory network (LSTM) to process time series features. The output features of these networks are then fused to form the final anomaly detection model.

[0171] Training data preparation: Prepare sample data of normal and abnormal behaviors for training the deep neural network model. Normal behavior samples can be data that has undergone privacy protection processing, while abnormal behavior samples can be generated by simulating data leakage or tampering.

[0172] Model training and optimization: Train a deep neural network model using training data, and adjust the model performance through cross-validation and hyperparameter optimization to ensure that the model can accurately detect abnormal behavior.

[0173] Step 3: Introduce an attention mechanism into the deep neural network model to weight key features in the multimodal feature vectors, thereby improving the model's detection accuracy for potential data leakage risks;

[0174] Specifically, attention mechanism design involves introducing an attention mechanism into the feature fusion layer of a deep neural network, such as using self-attention or channel attention. This attention mechanism automatically learns the importance weight of each feature in the feature vector.

[0175] Feature weighting: Based on the weights calculated by the attention mechanism, each feature in the multimodal feature vector is weighted to highlight key features (with weights higher than the first preset weight) and suppress unimportant features (with weights lower than the second preset weight).

[0176] Step 4: Improve detection accuracy: Input the weighted feature vector into the anomaly detection model to improve the model's accuracy in detecting abnormal behavior;

[0177] Specifically, the anomaly detection threshold setting is as follows: a threshold is set based on the anomaly score output by the model. When the anomaly score exceeds the threshold, it is judged as abnormal behavior.

[0178] Alarm mechanism triggering: When abnormal behavior is detected, the alarm mechanism is automatically triggered to notify the system administrator via email, SMS or system notification.

[0179] Anomaly information logging: Records detailed information about abnormal behavior, including the time of occurrence, data segment location, and abnormal characteristic values, to facilitate subsequent analysis and processing.

[0180] Step 5: When abnormal behavior is detected, the alarm mechanism is automatically triggered, and detailed information about the abnormal behavior is recorded, including the time of occurrence, the location of the data segment, and the abnormal characteristic value.

[0181] Based on the characteristics of abnormal behavior, privacy protection strategies can be dynamically adjusted, such as regenerating random matrices, increasing the noise intensity of obfuscated matrix data, or adjusting data storage paths to enhance the privacy protection of the data.

[0182] Specifically, abnormal behavior analysis involves analyzing the characteristics of abnormal behavior, such as the type of abnormality, frequency of occurrence, and scope of impact.

[0183] Privacy protection strategy adjustment: Select and adjust the appropriate privacy protection strategy based on the characteristics of abnormal behavior;

[0184] If data tampering is detected, a random matrix is ​​regenerated and the data is re-obfuscated.

[0185] If frequent access to data is detected, increase the noise level of the confusion matrix data;

[0186] If an attack is detected on the data storage path, adjust the data storage path and store the data fragments on a more secure node.

[0187] In some other embodiments of this specification, after evaluating the privacy protection strength of the confusion matrix data, the method further includes:

[0188] Step 1: Construct a data association graph based on graph neural networks (GNNs), where each data segment in the confusion matrix data is treated as a node in the graph, and the associations between data segments are treated as edges, forming a complex data association network.

[0189] Specifically, the data fragment node definition is as follows: the confusion matrix data is divided into multiple data fragments, and each data fragment is a node;

[0190] Edge definition based on relevance: Edges are defined based on the similarity or relevance between data segments. For example, relevance can be assessed by calculating the cosine similarity or mutual information between data segments;

[0191] Graph construction: Use graph neural networks (such as GCN or GAT) to construct data association graphs, where nodes represent data segments and edges represent the associations between data segments.

[0192] Step 2: Use graph neural networks to perform in-depth analysis of the data association graph, explore the potential relationships between data segments, and evaluate the importance of data segments in the graph.

[0193] In a data association graph, associations refer to the logical or semantic connections between data segments. These connections can be direct or indirect, reflecting the interdependence or similarity between the data segments. Direct associations: For example, two data segments may come from the same data source (such as the operation data and scheduling data of the same train), or there may be a clear causal relationship between them (such as equipment failure data and maintenance records). Indirect associations: Relationships indirectly connected through other data segments or nodes. For example, two data segments may be interconnected through multiple intermediate data segments (such as passenger flow data from multiple stations).

[0194] In graph neural networks, importance refers to the criticality of a data segment (node) within the entire graph, reflecting its role and value in the overall data structure. The norm of a node embedding vector is measured by the size (norm) of the node embedding vector learned through the graph neural network. A larger norm indicates a higher importance of the node in the graph. Node centrality metrics include degree centrality, betweenness centrality, and closeness centrality. These metrics measure the connectivity of a node in the network and its position on the critical path.

[0195] Specifically, graph neural network training: using graph neural networks to train data association graphs to learn the relationships between data segments;

[0196] Importance assessment: The importance of each data segment in the graph is evaluated using the output of the graph neural network. For example, the norm of a node's embedding vector or a node's centrality metric can be used to assess importance.

[0197] Potential Relationship Mining: By using the attention weights or edge prediction modules of graph neural networks, potential relationships between data segments can be mined.

[0198] Step 3: Based on the importance assessment results of the data association graph, perform layered encryption processing on the confusion matrix data. Use higher-level encryption algorithms for data segments with higher importance and use lighter-weight encryption algorithms for data segments with lower importance.

[0199] Specifically, encryption algorithm selection: Choose an appropriate encryption algorithm based on the importance of the data segment. For example, use AES-256 encryption algorithm for highly important data segments and AES-128 encryption algorithm for less important data segments;

[0200] Layered encryption implementation: Based on the importance assessment results, each data segment is subjected to layered encryption processing;

[0201] Encryption key management: Generate and manage encryption keys to ensure key security.

[0202] Step 4: During data storage and transmission, the layered encrypted data segments are dynamically reassembled. Based on the position and correlation of the data segments in the graph, the storage order and transmission path of the data segments are randomly adjusted to further reduce the risk of data leakage.

[0203] Specifically, the dynamic restructuring strategy is designed as follows:

[0204] Storage order adjustment: Based on the topology of nodes (data fragments) in the data association graph, the storage order of data fragments is randomly adjusted. For example, more important data fragments can be stored on more secure storage nodes, and their security can be enhanced through encryption and access control.

[0205] Transmission path adjustment: During data transmission, the transmission path is dynamically selected based on the correlation between data segments. For example, highly correlated data segments can be sent through different transmission paths to avoid simultaneous interception.

[0206] Introduction of randomness: Introducing randomness ensures that the order and path of storage and transmission are different each time, increasing the difficulty for attackers to analyze and crack the code.

[0207] The storage node selection is as follows:

[0208] Security assessment: Conduct a security assessment on all available storage nodes, including physical security, network security, and access control policies.

[0209] Dynamic allocation: Storage nodes are dynamically selected based on the importance of data fragments and security assessment results. For example, highly important data fragments are stored on high-security nodes, while less important data fragments are stored on ordinary nodes.

[0210] The transmission path plan is as follows:

[0211] Path diversity: Design multiple transmission paths to ensure that data fragments can be transmitted through different paths. Path selection can be based on factors such as network topology, bandwidth, and latency;

[0212] Random path selection: A path is randomly selected for each transmission, making the transmission path of data fragments unpredictable.

[0213] Step 5: When a usage instruction for rail transit data is received, the layered and encrypted data segments are quickly restored based on the structural information of the data association map to ensure the integrity and availability of the data.

[0214] Based on the same general inventive concept, this invention also protects a data leakage prevention processing device for rail transit based on matrix transformation, such as... Figure 2 As shown, Figure 2 This is a schematic diagram of the structure of the rail transit data leakage prevention processing device based on matrix transformation provided in an embodiment of the present invention. The following describes the rail transit data leakage prevention processing device based on matrix transformation provided by the present invention. The rail transit data leakage prevention processing device described below can be referred to in correspondence with the rail transit data leakage prevention processing method based on matrix transformation described above.

[0215] The data leakage prevention and control device for rail transit based on matrix transformation includes a matrix transformation module 201, a random matrix module 202, a matrix mixing module 203, a matrix evaluation module 204, and a matrix restoration module 205.

[0216] The matrix conversion module 201 is used to convert rail transit data into matrix data; where each row of the matrix data represents a data record and each column represents a data attribute.

[0217] The random matrix module 202 is used to generate an invertible random matrix based on the number of columns in the matrix data; wherein the number of rows and columns of the random matrix matches the number of columns in the matrix data.

[0218] Matrix mixing module 203 is used to multiply matrix data with a random matrix to obtain confused matrix data;

[0219] Matrix evaluation module 204 is used to evaluate the privacy protection strength of the obfuscated matrix data until the privacy protection strength meets the preset standard;

[0220] The matrix restoration module 205 is used to determine the inverse matrix of the random matrix when it receives the usage instruction of the rail transit data, and multiply the inverse matrix with the confusion matrix data to obtain the restored rail transit data.

[0221] Figure 3 This is a schematic diagram of the structure of the electronic device provided in an embodiment of the present invention.

[0222] like Figure 3 As shown, the electronic device may include a processor 310, a communications interface 320, a memory 330, and a communication bus 340. The processor 310, communications interface 320, and memory 330 communicate with each other via the communication bus 340. The processor 310 can call logical instructions from the memory 330 to execute a matrix transformation-based method for preventing the leakage of rail transit data.

[0223] Furthermore, the logical instructions in the aforementioned memory 330 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, essentially, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0224] On the other hand, the present invention also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer is able to execute the matrix transformation-based rail transit data leakage prevention processing method provided by the above methods.

[0225] In another aspect, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to perform the matrix transformation-based rail transit data leakage prevention processing method provided by the above methods.

[0226] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0227] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0228] Finally, 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 foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for preventing data leakage in rail transit based on matrix transformation, characterized in that, include: The rail transit data is converted into matrix data; wherein each row of the matrix data represents a data record and each column represents a data attribute. Based on the number of columns in the matrix data, an invertible random matrix is ​​generated; wherein the number of rows and columns of the random matrix matches the number of columns in the matrix data. Multiply the matrix data by the random matrix to obtain the confusion matrix data; Evaluate the privacy protection strength of the obfuscation matrix data until the privacy protection strength meets a preset standard; When a command to use rail transit data is received, the inverse matrix of the random matrix is ​​determined, and the inverse matrix is ​​multiplied by the confusion matrix data to obtain the restored rail transit data. Evaluate the privacy protection strength of the obfuscation matrix data until the privacy protection strength meets a preset standard, including: The assessment indicators for the strength of privacy protection are determined based on the characteristics of rail transit data and the needs for privacy protection. Calculate the privacy protection strength value of the confusion matrix data based on the evaluation indicators; Compare the privacy protection strength value with a preset standard; If the privacy protection strength value does not meet the preset standard, a random matrix is ​​regenerated, increasing the complexity of the random matrix, or noise is added to the confused matrix data to enhance the privacy protection strength value until the privacy protection strength meets the preset standard.

2. The method for preventing data leakage in rail transit based on matrix transformation according to claim 1, characterized in that, Converting rail transit data into matrix data includes: Determine the data type of each data attribute of the rail transit data, and use the mapping rule corresponding to the data type to map each data attribute to a column of matrix data; Arrange each data record according to its relationship to form a row of matrix data; The step of determining the data type of each data attribute of the rail transit data, and mapping each data attribute to a column of matrix data using a mapping rule corresponding to the data type, includes: When the data attribute is numerical, its value is directly used as a matrix element; When the data attributes are categorical, one-hot encoding is used to convert them into numerical data, which are then used as matrix elements. The correlation includes temporal order or spatial order.

3. The method for preventing data leakage in rail transit based on matrix transformation according to claim 1, characterized in that, Based on the number of columns in the matrix data, an invertible random matrix is ​​generated, including: Determine the dimensions of the random matrix so that its number of rows and columns matches the number of columns in the matrix data; Determine the distribution characteristics of rail transit data, and use a random generation algorithm corresponding to the distribution characteristics to generate elements of a random matrix; Check if a random matrix is ​​invertible by calculating its determinant or rank. If it is not invertible, regenerate it until an invertible random matrix is ​​obtained. The determination of the distribution characteristics of rail transit data, including generating elements of a random matrix using a random generation algorithm corresponding to the distribution characteristics, comprises: When the distribution characteristics of rail transit data are normally distributed, the element values ​​of the random matrix are generated using the normal distribution method. When the distribution characteristics of rail transit data are uniform, the element values ​​of the random matrix are generated using the uniform distribution method.

4. The method for preventing data leakage in rail transit based on matrix transformation according to claim 1, characterized in that, Multiplying the matrix data by the random matrix yields confused matrix data, including: The matrix size is determined based on the matrix data and the number of rows and columns of the random matrix; When the size of the matrix is ​​less than the minimum value of the preset size range, the matrix data is directly multiplied by the random matrix; When the size of the matrix falls within the preset size range, the matrix data is multiplied by the random matrix using either the Strassen algorithm or the Winograd algorithm. When the size of the matrix is ​​greater than the maximum value of the preset size range, the matrix data and the random matrix are divided into multiple sub-matrix blocks respectively; Perform matrix multiplication on each sub-matrix block to obtain the corresponding sub-result block; All sub-result blocks are concatenated according to their positions in the original matrix to form a complete confusion matrix.

5. The method for preventing data leakage in rail transit based on matrix transformation according to claim 1, characterized in that, The assessment indicators for the strength of privacy protection are determined based on the characteristics of rail transit data and privacy protection requirements, including: Assess the sensitivity of each data attribute in rail transit data, and determine the corresponding assessment index weights based on the level of sensitivity; For rail transit data where the proportion of categorical or numerical data exceeds a preset ratio, information entropy is selected as the evaluation index. The strength of privacy protection is quantified by calculating the probability distribution of each element value in the confusion matrix data. The higher the information entropy, the stronger the privacy protection. When it is necessary to assess the correlation between confusion matrix data and matrix data, mutual information is selected as the assessment metric; the lower the mutual information, the higher the strength of privacy protection. For rail transit data that needs to meet differential privacy protection requirements, the privacy budget value of differential privacy is selected as the evaluation index; the smaller the privacy budget value, the higher the privacy protection strength.

6. The method for preventing data leakage in rail transit based on matrix transformation according to claim 1, characterized in that, After evaluating the privacy protection strength of the obfuscation matrix data until the privacy protection strength meets a preset standard, the process further includes: The confusion matrix data is fragmented according to preset rules to generate multiple data segments; Each data segment is stored on a different storage node or storage medium to ensure isolation between data segments; When receiving instructions to use rail transit data, the following is also included: Collect all data fragments and perform restoration processing.

7. The method for preventing data leakage in rail transit based on matrix transformation according to claim 1, characterized in that, Before converting rail transit data into matrix data, the following steps are also included: Distinguish between structured and unstructured data in rail transit data; Combine the structured data with the unstructured data; The manner of combination includes at least one of the following: Feature concatenation, weighted averaging, or feature extraction and fusion from deep learning models.

8. The method for preventing data leakage in rail transit based on matrix transformation according to claim 1, characterized in that, After generating the invertible random matrix, the following steps are also included: A quantum entangled state generator is used to inject quantum entangled state information into each element of the random matrix; wherein the quantum entangled state information is represented by qubits, and the state of each qubit is determined by the element value of the random matrix; Before multiplying the matrix data with the random matrix, the method further includes: Quantum state verification is performed on the random matrix injected with quantum entangled states to ensure that the quantum state information of each element is complete and has not been tampered with.

9. A data leakage prevention and control device for rail transit based on matrix transformation, characterized in that, include: The matrix conversion module is used to convert rail transit data into matrix data; wherein each row of the matrix data represents a data record and each column represents a data attribute. A random matrix module is used to generate an invertible random matrix based on the number of columns in the matrix data; wherein the number of rows and columns of the random matrix matches the number of columns in the matrix data; A matrix mixing module is used to multiply the matrix data with the random matrix to obtain confused matrix data; The matrix evaluation module is used to evaluate the privacy protection strength of the obfuscated matrix data until the privacy protection strength meets a preset standard. The matrix restoration module is used to determine the inverse matrix of the random matrix when a usage instruction for rail transit data is received, and multiply the inverse matrix with the confusion matrix data to obtain the restored rail transit data. Evaluate the privacy protection strength of the obfuscation matrix data until the privacy protection strength meets a preset standard, including: The assessment indicators for the strength of privacy protection are determined based on the characteristics of rail transit data and the needs for privacy protection. Calculate the privacy protection strength value of the confusion matrix data based on the evaluation indicators; Compare the privacy protection strength value with a preset standard; If the privacy protection strength value does not meet the preset standard, a random matrix is ​​regenerated, increasing the complexity of the random matrix, or noise is added to the confused matrix data to enhance the privacy protection strength value until the privacy protection strength meets the preset standard.

Images