A big data-based payment platform transaction data processing system and method

Abstract

The application relates to the technical field of data processing, and discloses a payment platform transaction data processing system and method based on big data. The method collects historical transaction data in a payment platform, processes the historical transaction data through a data filtering mode, analyzes the filtered historical transaction data, clusters the filtered historical transaction data according to the analysis result, performs mutual exclusion analysis on the clustered historical transaction data, detects whether there is an inconsistency in the historical transaction data, processes the inconsistency in the historical transaction data, performs intrusion detection on the analyzed historical transaction data through a payment intrusion detection module constructed based on the analyzed historical transaction data, and classifies and processes according to the detection result, thereby improving the pertinence of payment platform transaction data processing.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, specifically to a payment platform transaction data processing system and method based on big data. Background Technology

[0002] With the continuous development of the electronic, digital and networked processes of commodity transactions, payment transaction platforms generate a large amount of transaction payment information as well as network attack information, which has caused a huge burden and interference to the security of transaction data on payment platforms.

[0003] The existing publicly disclosed patent application CN108469967A describes a method that, based on received transaction data, determines the transaction code corresponding to the transaction data; determines the memory address corresponding to the transaction code from the transaction array; and calls the dynamic library file pointed to by the memory address of the transaction code to process the transaction data. It can be seen that after each receipt of transaction data, the transaction code can be determined, the corresponding memory address can be determined from the transaction code, and finally, the dynamic library file pointed to by the memory address of the transaction code can be called to process the transaction data. Summary of the Invention

[0004] (a) Technical problems to be solved

[0005] To address the shortcomings of existing technologies, this invention provides a payment platform transaction data processing system and method based on big data, which has advantages such as security, accuracy, and efficiency, and solves the problem of heavy security burden on payment platform transaction data.

[0006] (II) Technical Solution:

[0007] To address the aforementioned technical problem of heavy security burden on payment platform transaction data, this invention provides the following technical solution:

[0008] This invention discloses a method for processing transaction data of a payment platform based on big data, specifically including the following steps:

[0009] S1. Collect historical transaction data from the payment platform and process the historical transaction data through data filtering to obtain filtered historical transaction data;

[0010] S2. Based on the analysis of the filtered historical transaction data, the filtered historical transaction data is clustered according to the analysis results, and the clustering results are output. The clustering results include: historical transaction data type, historical transaction data, and historical transaction data center.

[0011] S3. Perform mutual exclusion analysis on the clustered historical transaction data to detect any inconsistencies in the historical transaction data, process any inconsistencies, and output the analyzed historical transaction data.

[0012] S4. Construct a payment intrusion detection module. Use the constructed payment intrusion detection module to perform intrusion detection on the analyzed historical transaction data to obtain the detected historical transaction data.

[0013] S5. Set the data processing method to process the obtained historical transaction data after detection.

[0014] This invention collects historical transaction data from payment platforms and processes it through data filtering. It then analyzes the filtered historical transaction data, clusters it based on the analysis results, and performs mutual exclusion analysis on the clustered historical transaction data to detect and address any inconsistencies. Furthermore, based on the analyzed historical transaction data, a constructed payment intrusion detection module performs intrusion detection on the analyzed historical transaction data and categorizes and processes the data according to the detection results, thereby improving the targeted nature of payment platform transaction data processing.

[0015] Preferably, the process of collecting historical transaction data from the payment platform and processing the historical transaction data through data filtering to obtain filtered historical transaction data includes the following steps:

[0016] S11. Filter the collected historical transaction data;

[0017] The collected historical transaction data is iterated through, and the complete historical transaction data includes: the addresses of both parties, the order number, the transaction amount, the transaction time, and the transaction method;

[0018] Identify missing content in historical transaction data and delete any historical transaction data with missing content:

[0019] S12. The filtered historical transaction data is filtered a second time using a filtering algorithm.

[0020] Preferably, the secondary filtering of the filtered historical transaction data using a filtering algorithm includes the following steps:

[0021] Create a length of Select the array Each hash function calculates a hash for each piece of data in the historical transaction data and stores the results in an array, where... Indicates the number of hash functions. This indicates the number of data points in the collected historical transaction data.

[0022] During the hash function's calculation of historical transaction data, when two sets of data yield the same result, each piece of data in those two sets is compared.

[0023] When the comparison results are consistent, the two sets of data are set as identical.

[0024] When two sets of data are identical, compare the collection times of the two sets of data and retain the most recent set of data.

[0025] After the calculation is completed, the calculated historical transaction data is summarized to obtain the filtered historical transaction data.

[0026] This invention initially filters the collected historical transaction data by traversing it and identifying and deleting missing content. At the same time, it uses a filter algorithm to perform a secondary filter on the initially filtered historical transaction data, thereby improving the accuracy of historical transaction data processing.

[0027] Preferably, the step of analyzing the filtered historical transaction data and clustering the filtered historical transaction data based on the analysis results includes the following steps:

[0028] S21. Initialize the collected historical transaction data and construct the historical transaction dataset. ={ , ,..., };

[0029] in, Indicates the first Historical transaction data, This represents the first historical transaction data;

[0030] Select historical transaction data One set of historical transaction data as the initial historical transaction data center ={ , ,..., Each historical transaction data center is designated as representative data for a specific historical transaction data type.

[0031] S22. Cluster the collected historical transaction data;

[0032] Based on historical transaction data in the historical transaction dataset According to the distance formula, Calculate historical transaction data arrive The distance to each historical transaction data center is calculated, and the data center with the smallest distance is assigned to the corresponding class.

[0033] Set up to add 1 each time the calculation is completed, until all historical transaction data are clustered. Then use the result of this classification as the input for step S23, and then execute step S23.

[0034] in, Represents historical transaction data To the historical transaction data center distance, Indicates the first A historical transaction data center was set up. The value range is [1, ... ];

[0035] S23, Update historical transaction data center;

[0036] S24. Repeat steps S22-S23 until the historical transaction data in the historical transaction data center and category no longer changes, then output the historical transaction data type, historical transaction data, and historical transaction data center.

[0037] Preferably, the updated historical transaction data center includes the following steps:

[0038] Once all historical transaction data has been categorized, the historical transaction data center for each category will be recalculated based on the current categorization results. ;

[0039] in, This represents the distance from the r-th historical transaction data point in the category to the current historical transaction data point. This indicates the number of historical transaction data entries in the current category;

[0040] Based on the recalculation of historical transaction data for each category, the historical transaction data in each category is recalculated.

[0041] This invention improves the rationality of transaction data processing by initializing the collected historical transaction data, selecting initial cluster centers for clustering, and continuously updating the cluster centers and historical transaction data in the corresponding categories through iterative calculation until convergence.

[0042] Preferably, the step of performing mutual exclusion analysis on the clustered historical transaction data to detect inconsistencies, processing the inconsistencies, and outputting the analyzed historical transaction data includes the following steps:

[0043] S31. Iterate through the historical transaction data in each type of historical transaction data after clustering to determine whether there is a discrepancy between the transaction amount and the corresponding number of transactions.

[0044] S32. When it is detected that there is a transaction amount and a corresponding transaction quantity in the historical transaction data in the historical transaction data type, a transaction lock is set for each transaction item, and each transaction item is only allowed to be accessed by one set of transaction data during the transaction process;

[0045] S33. Recalculate the historical transaction data based on the set transaction lock, and summarize the calculation results to obtain the analyzed historical transaction data.

[0046] This invention improves the security of payment transactions by traversing historical transaction data in each type of historical transaction data after clustering, summarizing and comparing transaction amounts and corresponding transaction quantities, and processing and restricting abnormal transaction amounts and corresponding transaction quantities.

[0047] Preferably, the construction of the payment intrusion detection module involves performing intrusion detection on the analyzed historical transaction data using the constructed payment intrusion detection module. Obtaining the detected historical transaction data includes the following steps:

[0048] S41. Construct a payment intrusion detection module;

[0049] S42. Based on the construction of the payment intrusion detection module, intrusion detection is performed on the analyzed historical transaction data, and the detected historical transaction data is output.

[0050] Preferably, the construction of the payment intrusion detection module includes the following steps:

[0051] S411. Collect four types of payment transaction data to construct a training set, and extract features from the four types of payment transaction data in the training set;

[0052] The four types of payment transaction data include: flood attack payment transaction data, simulated attack payment transaction data, injection attack payment transaction data, and normal payment transaction data.

[0053] Preprocess the four types of payment transaction data in the training set;

[0054] Since some features of the four types of payment transaction data in the training set have real values ​​and some features have nominal attributes, before extracting features from the four types of payment transaction data, all nominal attributes are converted into integer values, missing values ​​are replaced with zeros, and deleted elements are replaced with redundant information.

[0055] The preprocessed four types of payment transaction data are aggregated and then input into a convolutional neural network for feature extraction.

[0056] After receiving the preprocessed four types of payment transaction data, the convolutional neural network moves through the input preprocessed four types of payment transaction data with the convolution kernel according to the set stride, and performs convolution calculation with the feature value of the corresponding region at each step, thereby realizing feature extraction of the preprocessed four types of payment transaction data;

[0057] The formula for calculating convolution is as follows:

[0058] ;

[0059] in, This represents the preprocessed four types of payment transaction data from the z-th input group. This represents the weights of the corresponding convolution kernel. b Indicates the bias value. Output characteristics of the four types of payment transaction data after preprocessing in group z;

[0060] S412. Train the features extracted by the convolutional neural network;

[0061] The features extracted by the convolutional neural network are used as the input of the backpropagation neural network;

[0062] The structure of a BP neural network includes: an input layer, hidden layers, and an output layer;

[0063] The learning algorithms for BP neural networks include:

[0064] Set the input of the input layer as ,in This represents the first feature of the input. This represents the m-th feature of the input, and the output of the output layer is... ,in This represents the first predicted feature in the output. Let v represent the s-th predicted feature of the output. The hidden layer contains q neurons, and v is the weight from the input layer to the hidden layer. The weights from the hidden layer to the output layer;

[0065] The training process of a BP neural network includes two stages: forward propagation of the signal and backward propagation of the error.

[0066] The forward propagation of a signal is that data is input from the input layer, passes through the hidden layer, and finally reaches the output layer;

[0067] The formula for calculating the forward propagation of a signal is:

[0068] ;

[0069] ;

[0070] in, This represents the input to the h-th hidden layer neuron. The weight represents the weight from the i-th input in the input layer to the h-th hidden layer neuron, and m represents the number of input data in the input layer; This represents the i-th input of the input layer; This represents the bias value of the h-th hidden layer neuron; This represents the input to the j-th output layer neuron. q represents the weight from the h-th hidden layer neuron to the j-th output layer neuron, and q represents the number of hidden layer neurons; This represents the bias value of the j-th output layer neuron. This represents the output of the j-th output layer neuron;

[0071] Backpropagation stage of error:

[0072] Calculate the error between the output layer and the expected value, set an error threshold between the output layer and the expected value, and adjust the weights through the backpropagation stage of the error when the error between the output layer and the expected value is greater than or equal to the set error threshold.

[0073] The error calculation formula is as follows:

[0074] ;

[0075] Where E is the error, Let be the expected value of the j-th output, and n be the number of output neurons;

[0076] The formula for calculating the weight adjustment is as follows:

[0077] ;

[0078] in, Indicates the weighted adjustment value. l Let E represent the learning rate, E represent the error, and Y represent the output of the output layer.

[0079] The weights are adjusted by updating the current weights with weight adjustment values. The adjusted weights are then used as new weights for the forward propagation calculation of the signal and iteratively calculated until a definite BP neural network model is obtained.

[0080] The payment intrusion detection module is defined as a combination of a BP neural network model and a convolutional neural network, where the convolutional neural network is the front end of the payment intrusion detection module and the BP neural network model is the back end of the payment intrusion detection module;

[0081] S42. Based on the construction of a payment intrusion detection module, perform intrusion detection on the analyzed historical transaction data and output the detected historical transaction data.

[0082] The analyzed historical transaction data is input into the constructed payment intrusion detection module. The features of the analyzed historical transaction data are extracted, and the BP neural network model in the payment intrusion detection module is used to detect the extracted historical transaction data features, determine the type of payment transaction data to which the analyzed historical transaction data belongs, and output the detected historical transaction data.

[0083] This invention constructs a payment intrusion detection module by collecting four types of payment transaction data, extracting features from the collected data, and iteratively training the data. Simultaneously, based on this module, intrusion detection is performed on analyzed historical transaction data to assess the attack status of payment transaction data in real time, thereby improving the security of payment transactions.

[0084] Preferably, the data processing method for processing the obtained historical transaction data after detection includes the following steps:

[0085] Process the data according to the type of payment transaction it belongs to in the historical transaction data;

[0086] To prevent flooding attacks, set access permissions for payment transaction data and isolate the attack.

[0087] For simulated attack payment transaction data, a private key is used to encrypt the payment transaction source address;

[0088] To prevent injection attacks, a user verification method should be set up for payment transaction data to isolate the injection attack.

[0089] The present invention also discloses a payment platform transaction data processing system based on big data, which is used to implement a payment platform transaction data processing method based on big data. The system includes: a data acquisition module, a data processing module, a mutual exclusion analysis module, a clustering module, a payment intrusion detection module, and a transaction data processing module.

[0090] The data acquisition module is used to collect historical transaction data from the payment platform;

[0091] The data processing module is used to process historical transaction data through data filtering.

[0092] The clustering module is used to cluster the processed historical transaction data using a clustering method.

[0093] The mutual exclusion analysis module is used to perform mutual exclusion analysis on the clustered historical transaction data.

[0094] The payment intrusion detection module is used to perform payment intrusion detection on historical transaction data after mutual exclusion analysis.

[0095] The transaction data processing module is used to perform specific processing based on the payment intrusion detection results.

[0096] (III) Beneficial Effects:

[0097] Compared with existing technologies, this invention provides a payment platform transaction data processing system and method based on big data, which has the following beneficial effects:

[0098] 1. This invention collects historical transaction data from payment platforms and processes it through data filtering. It then analyzes the filtered historical transaction data, clusters it based on the analysis results, and performs mutual exclusion analysis on the clustered historical transaction data to detect inconsistencies and address them. Furthermore, based on the analyzed historical transaction data, a constructed payment intrusion detection module performs intrusion detection on the analyzed historical transaction data and classifies and processes it according to the detection results, thereby improving the targeted nature of payment platform transaction data processing.

[0099] 2. This invention initially filters the collected historical transaction data by traversing the historical transaction data and identifying and deleting missing content in the historical transaction data. At the same time, a filter algorithm is used to perform a secondary filter on the initially filtered historical transaction data, thereby improving the accuracy of historical transaction data processing.

[0100] 3. This invention improves the rationality of transaction data processing by initializing the collected historical transaction data, selecting initial cluster centers for clustering, and continuously updating the cluster centers and corresponding historical transaction data in the corresponding categories through iterative calculation until convergence.

[0101] 4. This invention improves the security of payment transactions by traversing the historical transaction data in each type of historical transaction data after clustering, summarizing and comparing the transaction amount and the corresponding number of transactions, and processing and restricting abnormal transaction amounts and corresponding number of transactions.

[0102] 5. This invention constructs a payment intrusion detection module by collecting four types of payment transaction data and extracting features from the collected four types of payment transaction data and iteratively training the data. At the same time, based on the construction of the payment intrusion detection module, intrusion detection is performed on the analyzed historical transaction data, and the attack status of payment transaction data is judged in real time, thereby improving the security of payment transactions. Attached Figure Description

[0103] Figure 1 This is a schematic diagram of the transaction data processing flow structure of the payment platform of the present invention. Detailed Implementation

[0104] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0105] Example 1: Please refer to Figure 1 This embodiment discloses a method for processing transaction data of a payment platform based on big data, specifically including the following steps:

[0106] S1. Collect historical transaction data from the payment platform and process the historical transaction data through data filtering to obtain filtered historical transaction data;

[0107] S2. Based on the analysis of the filtered historical transaction data, the filtered historical transaction data is clustered according to the analysis results, and the clustering results are output. The clustering results include: historical transaction data type, historical transaction data, and historical transaction data center.

[0108] S3. Perform mutual exclusion analysis on the clustered historical transaction data to detect any inconsistencies in the historical transaction data, process any inconsistencies, and output the analyzed historical transaction data.

[0109] S4. Construct a payment intrusion detection module. Use the constructed payment intrusion detection module to perform intrusion detection on the analyzed historical transaction data to obtain the detected historical transaction data.

[0110] S5. Set the data processing method to process the obtained historical transaction data after detection;

[0111] Further, please refer to Figure 1 Collect historical transaction data from payment platforms and process the historical transaction data through data filtering to obtain filtered historical transaction data. The steps include:

[0112] S11. Filter the collected historical transaction data;

[0113] The collected historical transaction data is iterated through, and the complete historical transaction data includes: the addresses of both parties, the order number, the transaction amount, the transaction time, and the transaction method;

[0114] Identify missing content in historical transaction data and delete any historical transaction data with missing content:

[0115] S12. Perform secondary filtering on the filtered historical transaction data using a filtering algorithm;

[0116] Create a length of Select the array Each hash function calculates a hash for each piece of data in the historical transaction data and stores the results in an array, where... Indicates the number of hash functions. This indicates the number of data points in the collected historical transaction data.

[0117] During the hash function's calculation of historical transaction data, when two sets of data yield the same result, each piece of data in those two sets is compared.

[0118] When the comparison results are consistent, the two sets of data are set as identical.

[0119] When two sets of data are identical, compare the collection times of the two sets of data and retain the most recent set of data.

[0120] After the calculation is completed, the calculated historical transaction data is summarized to obtain the filtered historical transaction data;

[0121] Further, please refer to Figure 1 Based on the analysis of the filtered historical transaction data, the filtered historical transaction data is clustered according to the analysis results, and the clustering results are output. The clustering results include: historical transaction data type, historical transaction data, and historical transaction data center. The steps include:

[0122] S21. Initialize the collected historical transaction data and construct the historical transaction dataset. ={ , ,..., };

[0123] in, Indicates the first Historical transaction data, This represents the first historical transaction data;

[0124] Select historical transaction data One set of historical transaction data as the initial historical transaction data center ={ , ,..., Each historical transaction data center is designated as representative data for a specific historical transaction data type.

[0125] S22. Cluster the collected historical transaction data;

[0126] Based on historical transaction data in the historical transaction dataset According to the distance formula, Calculate historical transaction data arrive The distance to each historical transaction data center is calculated, and the data center with the smallest distance is assigned to the corresponding class.

[0127] Furthermore, we set it to add one at a time for each calculation until all historical transaction data are clustered. Then, we use the result of this classification as the input for step S23 and execute step S23.

[0128] in, Represents historical transaction data To the historical transaction data center distance, Indicates the first A historical transaction data center was set up. The value range is [1, ... ];

[0129] S23, Update historical transaction data center;

[0130] Once all historical transaction data has been categorized, the historical transaction data center for each category will be recalculated based on the current categorization results. ;

[0131] in, This represents the distance from the r-th historical transaction data point in the category to the current historical transaction data point. This indicates the number of historical transaction data entries in the current category;

[0132] Based on recalculating the historical transaction data center for each category, the historical transaction data for each category is recalculated;

[0133] S24. Repeat steps S22-S23 until the historical transaction data in the historical transaction data center and category no longer changes, then output the historical transaction data type, historical transaction data, and historical transaction data center.

[0134] Further, please refer to Figure 1 The process involves performing mutual exclusion analysis on the clustered historical transaction data to detect inconsistencies, handling these inconsistencies, and outputting the analyzed historical transaction data, including the following steps:

[0135] S31. Iterate through the historical transaction data in each type of historical transaction data after clustering to determine whether there is a discrepancy between the transaction amount and the corresponding number of transactions.

[0136] S32. When it is detected that there is a transaction amount and a corresponding transaction quantity in the historical transaction data in the historical transaction data type, a transaction lock is set for each transaction item, and each transaction item is only allowed to be accessed by one set of transaction data during the transaction process;

[0137] S33. Recalculate the historical transaction data based on the set transaction lock, and summarize the calculation results to obtain the analyzed historical transaction data;

[0138] Further, please refer to Figure 1 A payment intrusion detection module is constructed. This module is then used to perform intrusion detection on the analyzed historical transaction data. The obtained historical transaction data includes the following steps:

[0139] S41. Construct a payment intrusion detection module;

[0140] S411. Collect four types of payment transaction data to construct a training set, and extract features from the four types of payment transaction data in the training set;

[0141] The four types of payment transaction data include: flood attack payment transaction data, simulated attack payment transaction data, injection attack payment transaction data, and normal payment transaction data.

[0142] Furthermore, the four types of payment transaction data in the training set are preprocessed;

[0143] Since some features of the four types of payment transaction data in the training set have real values ​​and some features have nominal attributes, before extracting features from the four types of payment transaction data, all nominal attributes are converted into integer values, missing values ​​are replaced with zeros, and deleted elements are replaced with redundant information.

[0144] The preprocessed four types of payment transaction data are aggregated and then input into a convolutional neural network for feature extraction.

[0145] After receiving the preprocessed four types of payment transaction data, the convolutional neural network moves through the input preprocessed four types of payment transaction data with the convolution kernel according to the set stride, and performs convolution calculation with the feature value of the corresponding region at each step, thereby realizing feature extraction of the preprocessed four types of payment transaction data;

[0146] The formula for calculating convolution is as follows:

[0147] ;

[0148] in, This represents the preprocessed four types of payment transaction data from the z-th input group. This represents the weights of the corresponding convolution kernel. b Indicates the bias value. Output characteristics of the four types of payment transaction data after preprocessing in group z;

[0149] S412. Train the features extracted by the convolutional neural network;

[0150] The features extracted by the convolutional neural network are used as the input of the backpropagation neural network;

[0151] The structure of a BP neural network includes: an input layer, hidden layers, and an output layer;

[0152] The learning algorithms for BP neural networks include:

[0153] Set the input of the input layer as ,in This represents the first feature of the input. This represents the m-th feature of the input, and the output of the output layer is... ,in This represents the first predicted feature in the output. Let v represent the s-th predicted feature of the output. The hidden layer contains q neurons, and v is the weight from the input layer to the hidden layer. The weights from the hidden layer to the output layer;

[0154] The training process of a BP neural network includes two stages: forward propagation of the signal and backward propagation of the error.

[0155] The forward propagation of a signal is that data is input from the input layer, passes through the hidden layer, and finally reaches the output layer;

[0156] The formula for calculating the forward propagation of a signal is:

[0157] ;

[0158] ;

[0159] in, This represents the input to the h-th hidden layer neuron. The weight represents the weight from the i-th input in the input layer to the h-th hidden layer neuron, and m represents the number of input data in the input layer; This represents the i-th input of the input layer; This represents the bias value of the h-th hidden layer neuron; This represents the input to the j-th output layer neuron. q represents the weight from the h-th hidden layer neuron to the j-th output layer neuron, and q represents the number of hidden layer neurons; This represents the bias value of the j-th output layer neuron. This represents the output of the j-th output layer neuron;

[0160] Backpropagation stage of error:

[0161] Calculate the error between the output layer and the expected value, set an error threshold between the output layer and the expected value, and adjust the weights through the backpropagation stage of the error when the error between the output layer and the expected value is greater than or equal to the set error threshold.

[0162] The error calculation formula is as follows:

[0163] ;

[0164] Where E is the error, Let be the expected value of the j-th output, and n be the number of output neurons;

[0165] The formula for calculating the weight adjustment is as follows:

[0166] ;

[0167] in, Indicates the weighted adjustment value. l Let E represent the learning rate, E represent the error, and Y represent the output of the output layer.

[0168] The weights are adjusted by updating the current weights with weight adjustment values. The adjusted weights are then used as new weights for the forward propagation calculation of the signal and iteratively calculated until a definite BP neural network model is obtained.

[0169] Furthermore, the payment intrusion detection module is defined as a combination of a BP neural network model and a convolutional neural network, where the convolutional neural network is the front end of the payment intrusion detection module and the BP neural network model is the back end of the payment intrusion detection module.

[0170] S42. Based on the construction of a payment intrusion detection module, perform intrusion detection on the analyzed historical transaction data and output the detected historical transaction data.

[0171] The analyzed historical transaction data is input into the constructed payment intrusion detection module, the features of the analyzed historical transaction data are extracted, and the BP neural network model in the payment intrusion detection module is used to detect the extracted historical transaction data features, determine the type of payment transaction data to which the analyzed historical transaction data belongs, and output the detected historical transaction data.

[0172] Further, please refer to Figure 1 The process of processing the obtained historical transaction data after detection includes the following steps:

[0173] Process the data according to the type of payment transaction it belongs to in the historical transaction data;

[0174] To prevent flooding attacks, set access permissions for payment transaction data and isolate the attack.

[0175] For simulated attack payment transaction data, a private key is used to encrypt the payment transaction source address;

[0176] To prevent injection attacks, a user verification method should be set up for payment transaction data to isolate the injection attack.

[0177] Example 2: Please refer to Figure 1 This embodiment also discloses a payment platform transaction data processing system based on big data, used to implement a payment platform transaction data processing method based on big data. The system includes: a data acquisition module, a data processing module, a mutual exclusion analysis module, a clustering module, a payment intrusion detection module, and a transaction data processing module.

[0178] The data acquisition module is used to collect historical transaction data from the payment platform;

[0179] The data processing module is used to process historical transaction data through data filtering.

[0180] The clustering module is used to cluster the processed historical transaction data using a clustering method.

[0181] The mutual exclusion analysis module is used to perform mutual exclusion analysis on the clustered historical transaction data.

[0182] The payment intrusion detection module is used to perform payment intrusion detection on historical transaction data after mutual exclusion analysis.

[0183] The transaction data processing module is used to perform specific processing based on the payment intrusion detection results.

[0184] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for processing transaction data of a payment platform based on big data, characterized in that, Includes the following steps: S1. Collect historical transaction data from the payment platform and process the historical transaction data through data filtering to obtain filtered historical transaction data; S2. Based on the analysis of the filtered historical transaction data, the filtered historical transaction data is clustered according to the analysis results, and the clustering results are output. The clustering results include: historical transaction data type, historical transaction data, and historical transaction data center. S3. Perform mutual exclusion analysis on the clustered historical transaction data to detect any inconsistencies in the historical transaction data, process any inconsistencies, and output the analyzed historical transaction data. S3 includes the following steps: S31. Iterate through the historical transaction data in each type of historical transaction data after clustering to determine whether there is a discrepancy between the transaction amount and the corresponding number of transactions. S32. When it is detected that there is a transaction amount and a corresponding transaction quantity in the historical transaction data in the historical transaction data type, a transaction lock is set for each transaction item, and each transaction item is only allowed to be accessed by one set of transaction data during the transaction process; S33. Recalculate the historical transaction data based on the set transaction lock, and summarize the calculation results to obtain the analyzed historical transaction data; S4. Construct a payment intrusion detection module. Use the constructed payment intrusion detection module to perform intrusion detection on the analyzed historical transaction data to obtain the detected historical transaction data. S4 includes the following steps: S41. Construct a payment intrusion detection module, including the following steps: Collect four types of payment transaction data to construct a training set, and extract features from the four types of payment transaction data in the training set; The four types of payment transaction data include flood attack payment transaction data, simulated attack payment transaction data, injection attack payment transaction data, and normal payment transaction data. Preprocess the four types of payment transaction data in the training set; The preprocessed four types of payment transaction data are aggregated and then input into a convolutional neural network for feature extraction. After receiving the preprocessed four types of payment transaction data, the convolutional neural network moves through the input preprocessed four types of payment transaction data with the convolutional kernel according to the set stride, and performs convolution calculation with the feature value of the corresponding region at each step, thereby realizing feature extraction of the preprocessed four types of payment transaction data; Train the convolutional neural network on the features extracted by the convolutional neural network; The features extracted by the convolutional neural network are used as the input of the backpropagation neural network; The payment intrusion detection module is defined as a combination of a BP neural network model and a convolutional neural network, where the convolutional neural network is the front end of the payment intrusion detection module and the BP neural network model is the back end of the payment intrusion detection module; S42. Based on the payment intrusion detection module, perform intrusion detection on the analyzed historical transaction data and output the detected historical transaction data, including the following steps: The analyzed historical transaction data is input into the constructed payment intrusion detection module, the features of the analyzed historical transaction data are extracted, and the BP neural network model in the payment intrusion detection module is used to detect the extracted historical transaction data features, determine the type of payment transaction data to which the analyzed historical transaction data belongs, and output the detected historical transaction data. S5. Set the data processing method to process the obtained historical transaction data after detection.

2. The method for processing transaction data of a payment platform based on big data according to claim 1, characterized in that, The process of collecting historical transaction data from the payment platform and processing the historical transaction data through data filtering to obtain filtered historical transaction data includes the following steps: S11. Filter the collected historical transaction data; The collected historical transaction data is iterated through, and the complete historical transaction data includes: the addresses of both parties, the order number, the transaction amount, the transaction time, and the transaction method; Identify missing content in historical transaction data and delete any historical transaction data with missing content: S12. The filtered historical transaction data is filtered a second time using a filtering algorithm.

3. The method for processing transaction data of a payment platform based on big data according to claim 2, characterized in that, The secondary filtering of the filtered historical transaction data using a filtering algorithm includes the following steps: Create a length of Select the array Each hash function calculates a hash for each piece of data in the historical transaction data and stores the results in an array, where... Indicates the number of hash functions. This indicates the number of data points in the collected historical transaction data. During the hash function's calculation of historical transaction data, when two sets of data yield the same result, each piece of data in those two sets is compared. When the comparison results are consistent, the two sets of data are set as identical. When two sets of data are identical, compare the collection times of the two sets of data and retain the most recent set of data. After the calculation is completed, the calculated historical transaction data is summarized to obtain the filtered historical transaction data.

4. The method for processing transaction data of a payment platform based on big data according to claim 1, characterized in that, The step of analyzing the filtered historical transaction data and clustering the filtered historical transaction data based on the analysis results includes the following steps: S21. Initialize the collected historical transaction data and construct the historical transaction dataset. ={ , ,..., }; in, Indicates the first Historical transaction data, This represents the first historical transaction data; Select historical transaction data One set of historical transaction data as the initial historical transaction data center ={ , ,..., Each historical transaction data center is defined as representative data for a specific historical transaction data type. S22. Cluster the collected historical transaction data; Based on historical transaction data in the historical transaction dataset According to the distance formula, Calculate historical transaction data arrive The distance to each historical transaction data center is calculated, and the data center with the smallest distance is assigned to the corresponding class. Set up to add 1 each time the calculation is completed, until all historical transaction data are clustered. Then use the result of this classification as the input for step S23, and then execute step S23. in, Represents historical transaction data To the historical transaction data center distance, Indicates the first A historical transaction data center was set up. The value range is [1, ... ]; S23, Update historical transaction data center; S24. Repeat steps S22-S23 until the historical transaction data in the historical transaction data center and category no longer changes, then output the historical transaction data type, historical transaction data, and historical transaction data center.

5. The method for processing transaction data of a payment platform based on big data according to claim 4, characterized in that, The process of updating the historical transaction data center includes the following steps: Once all historical transaction data has been categorized, the historical transaction data center for each category will be recalculated based on the current categorization results. ; in, This represents the distance from the r-th historical transaction data point in the category to the current historical transaction data point. This indicates the number of historical transaction data entries in the current category; Based on the recalculation of historical transaction data for each category, the historical transaction data in each category is recalculated.

6. The method for processing transaction data of a payment platform based on big data according to claim 1, characterized in that, The construction of the payment intrusion detection module also includes the following steps: The training process of a BP neural network includes two stages: forward propagation of the signal and backward propagation of the error. The formula for calculating the forward propagation of a signal is: ； ； in, This represents the input to the h-th hidden layer neuron. The weight represents the weight from the i-th input in the input layer to the h-th hidden layer neuron, and m represents the number of input data in the input layer; This represents the i-th input of the input layer; This represents the bias value of the h-th hidden layer neuron; This represents the input to the j-th output layer neuron. q represents the weight from the h-th hidden layer neuron to the j-th output layer neuron, and q represents the number of hidden layer neurons; This represents the bias value of the j-th output layer neuron. This represents the output of the j-th output layer neuron; Backpropagation stage of error: Calculate the error between the output layer and the expected value, set an error threshold between the output layer and the expected value, and adjust the weights through the backpropagation stage of the error when the error between the output layer and the expected value is greater than or equal to the set error threshold. The weights are adjusted by updating the current weights with weight adjustment values. The adjusted weights are then used as new weights for the forward propagation calculation of the signal and the calculation is performed iteratively until a definite BP neural network model is obtained.

7. A system for implementing the big data-based payment platform transaction data processing method according to any one of claims 1-6, characterized in that, include: The system includes a data acquisition module, a data processing module, a mutual exclusion analysis module, a clustering module, a payment intrusion detection module, and a transaction data processing module. The data acquisition module is used to collect historical transaction data from the payment platform; The data processing module is used to process historical transaction data through data filtering. The clustering module is used to cluster the processed historical transaction data using a clustering method. The mutual exclusion analysis module is used to perform mutual exclusion analysis on the clustered historical transaction data. The payment intrusion detection module is used to perform payment intrusion detection on historical transaction data after mutual exclusion analysis. The transaction data processing module is used to perform specific processing based on the payment intrusion detection results.

Images