Financial fraud detection method and device, electronic equipment and medium
By acquiring multiple financial data and generating comprehensive feature vectors, and using pre-trained fraud detection models, the problem of inefficient traditional financial fraud detection is solved, and the rapid and accurate identification of fraudulent behaviors of audio and video synthesis is achieved.
Patent Information
- Application Number
- CN202510531861.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-04-25
- Publication Date
- 2025-08-05
AI Technical Summary
Traditional financial fraud detection methods are inefficient and difficult to identify complex and hidden fraudulent behaviors, especially fraudulent behaviors that are made using AI audio and video synthesis technology are difficult to detect.
Financial data of multiple dimensions are obtained, including transaction data, user behavior data and audio and video data, target features are extracted through feature extraction module, comprehensive feature vectors are generated, and pre-trained fraud detection model is used for detection.
It realizes rapid and accurate identification of financial fraud, improves detection efficiency and intelligence, and can effectively identify audio and video synthesis fraud.
Smart Images
Figure CN120429610A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of information processing, and more specifically, to a financial fraud detection method, device, electronic device, and medium. Background Art
[0002] With the development of internet technology, fraud is becoming increasingly common and complex in the financial sector. Traditional fraud detection methods rely primarily on rule-based matching and manual review. These methods are inefficient when processing large amounts of data and struggle to identify complex and covert fraudulent activity. Fraudulent activity using AI-powered audio and video synthesis, in particular, poses significant risks and challenges to financial institutions due to its high degree of realism and difficulty in detection. Therefore, a more intelligent and efficient fraud detection approach is needed to address this challenge. Summary of the Invention
[0003] In view of this, the purpose of this application is to provide a financial fraud detection method, device, electronic device and medium that can detect synthesized audio and video and quickly and accurately identify financial fraud behaviors.
[0004] An embodiment of the present application provides a method for detecting financial fraud, the method comprising the following steps:
[0005] Obtaining financial data of target users in multiple dimensions; the financial data includes transaction data, user behavior data, and audio and video data; the audio and video data is the audio and video evidence provided by the user during the application and / or transaction process;
[0006] Through the feature extraction module that matches the financial data type, the target features of each type of financial data are extracted respectively, and the target features of multiple dimensions are obtained;
[0007] Fusing the target features of the multiple dimensions to generate a comprehensive feature vector;
[0008] The comprehensive feature vector is processed by a pre-trained fraud detection model to detect whether fraudulent behavior exists and output a fraud detection result.
[0009] In some embodiments, in the financial fraud detection method, target features of each type of financial data are extracted using a feature extraction module that matches the type of financial data, thereby obtaining target features of multiple dimensions, including:
[0010] Target fraud features of transaction data and user behavior data are extracted using statistical analysis and machine learning algorithms. These target fraud features represent characteristics related to fraudulent behavior.
[0011] For audio and video data, audio and video synthesis features are extracted as the target features of this dimension.
[0012] In some embodiments, in the financial fraud detection method, extracting audio and video synthesis features from audio and video data includes:
[0013] For audio and video data, extract audio tampering trace features and video frame repetition rate features;
[0014] The audio and video synthesis feature is generated based on the audio tampering trace feature and the video frame repetition rate feature.
[0015] In some embodiments, in the financial fraud detection method, after outputting the fraud detection result, the method further includes:
[0016] When the fraud detection result indicates that fraud exists, extracting target attribute information representing attributes of the transaction behavior from the transaction information;
[0017] processing the comprehensive feature vector to generate target fraud evidence information;
[0018] A target fraud report is generated based on a predefined fraud report template, the target attribute information, and the target fraud evidence information.
[0019] In some embodiments, in the financial fraud detection method, processing the comprehensive feature vector to generate target fraud evidence information includes at least one of the following:
[0020] Processing the audio and video synthesis features in the comprehensive feature vector to generate audio and video fraud evidence;
[0021] The target fraud features in the comprehensive feature vector are processed to generate quantified behavioral anomaly evidence.
[0022] In some embodiments, in the financial fraud detection method, after generating the target fraud report, the method further includes:
[0023] When the fraud detection result indicates that fraud exists, generating fraud warning information;
[0024] Obtaining target countermeasure information matching the fraudulent behavior detection result from pre-configured countermeasure information;
[0025] The target fraud report, fraud warning information, target response information and fraud behavior detection results are sent to the target terminal device.
[0026] In some embodiments, in the financial fraud detection method, sending the target fraud report, fraud warning information, target response information, and fraud behavior detection results to the target terminal device includes:
[0027] Based on different pre-configured roles of the target terminal device, the target fraud report and target countermeasure information are processed to obtain the target fraud report and target countermeasure information that match the role; different roles have different permissions;
[0028] The fraud warning information, fraud behavior detection results, target fraud report matching the role, and target response measures information are sent to the target terminal device corresponding to the role.
[0029] In some embodiments, a financial fraud detection device is further provided, the device comprising:
[0030] An acquisition module is used to obtain financial data of target users in multiple dimensions; the financial data includes transaction data, user behavior data, and audio and video data; the audio and video data is the audio and video evidence provided by the user during the application and / or transaction process;
[0031] An extraction module is used to extract target features of each type of financial data through a feature extraction module that matches the type of financial data, thereby obtaining target features of multiple dimensions;
[0032] A fusion module, configured to fuse the target features of the multiple dimensions to generate a comprehensive feature vector;
[0033] The detection module is used to process the comprehensive feature vector through a pre-trained fraud detection model to detect whether fraudulent behavior exists and output a fraud detection result.
[0034] In some embodiments, an electronic device is also provided, comprising: a processor, a memory, and a bus, wherein the memory stores machine-readable instructions executable by the processor. When the electronic device is running, the processor and the memory communicate through the bus, and when the machine-readable instructions are executed by the processor, the steps of the financial fraud detection method are performed.
[0035] In some embodiments, a computer-readable storage medium is further provided, on which a computer program is stored. When the computer program is executed by a processor, the steps of the financial fraud detection method are executed.
[0036] In an embodiment of the present application, a financial fraud detection method, apparatus, electronic device, and medium are provided. The method obtains financial data of multiple dimensions of a target user; the financial data includes transaction data, user behavior data, and audio and video data; the audio and video data is audio and video evidence provided by the user during the application and / or transaction process; a feature extraction module that matches the type of financial data is used to extract target features of each type of financial data to obtain target features of multiple dimensions; the target features of the multiple dimensions are integrated to generate a comprehensive feature vector; the comprehensive feature vector is processed by a pre-trained fraud detection model to detect whether fraudulent behavior exists and output a fraud detection result, thereby combining the audio and video data to determine whether this behavior is synthesized to more accurately detect financial fraud; and real-time detection by the fraud detection model can more intelligently and efficiently detect financial fraud. BRIEF DESCRIPTION OF THE DRAWINGS
[0037] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following is a brief introduction to the drawings required for use in the embodiments. It should be understood that the following drawings only show certain embodiments of the present application and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other relevant drawings can be obtained based on these drawings without creative work.
[0038] Figure 1 A flowchart of the financial fraud detection method according to an embodiment of the present application is shown;
[0039] Figure 2 A flowchart of a method for extracting target features of each type of financial data by using a feature extraction module matched to the type of financial data according to an embodiment of the present application is shown;
[0040] Figure 3 A flowchart of another financial fraud detection method according to an embodiment of the present application is shown;
[0041] Figure 4 A flowchart of another financial fraud detection method according to an embodiment of the present application is shown;
[0042] Figure 5 The following is a structural block diagram of the financial fraud detection device according to an embodiment of the present application;
[0043] Figure 6 The figure shows a structural block diagram of the electronic device described in the embodiment of the present application. DETAILED DESCRIPTION
[0044] In order to make the purpose, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. It should be understood that the drawings in the present application only serve the purpose of illustration and description and are not used to limit the scope of protection of the present application. In addition, it should be understood that the schematic drawings are not drawn to scale. The flowcharts used in this application illustrate the operations implemented according to some embodiments of the present application. It should be understood that the operations of the flowcharts can be implemented out of sequence, and steps without logical context can be reversed or implemented simultaneously. In addition, those skilled in the art, under the guidance of the contents of this application, can add one or more other operations to the flowchart, or remove one or more operations from the flowchart.
[0045] In addition, the described embodiments are only a part of the embodiments of the present application, rather than all of the embodiments. The components of the embodiments of the present application generally described and shown in the drawings here can be arranged and designed in various configurations. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the claimed application, but merely represents selected embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without making creative work are within the scope of protection of the present application.
[0046] It should be noted that the term "comprising" will be used in the embodiments of the present application to indicate the existence of the features declared thereafter, but does not exclude the addition of other features.
[0047] With the development of internet technology, fraud is becoming increasingly common and complex in the financial sector. Traditional fraud detection methods rely primarily on rule-based matching and manual review. These methods are inefficient when processing large amounts of data and struggle to identify complex and covert fraudulent activity. Fraudulent activity using AI-powered audio and video synthesis, in particular, poses significant risks and challenges to financial institutions due to its high degree of realism and difficulty in detection. Therefore, a more intelligent and efficient fraud detection approach is needed to address this challenge.
[0048] Based on this, in an embodiment of the present application, a financial fraud detection method, device, electronic device and medium are provided, wherein the method obtains financial data of multiple dimensions of a target user; the financial data includes transaction data, user behavior data and audio and video data; the audio and video data is audio and video evidence provided by the user during the application and / or transaction process; the target features of each type of financial data are extracted separately by a feature extraction module that matches the type of financial data to obtain target features of multiple dimensions; the target features of the multiple dimensions are integrated to generate a comprehensive feature vector; the comprehensive feature vector is processed by a pre-trained fraud detection model to detect whether there is fraudulent behavior, and the fraud behavior detection result is output, thereby combining whether the audio and video data synthesizes this behavior to more accurately detect financial fraud; real-time detection is performed by the fraud detection model, which can detect financial fraud more intelligently and efficiently.
[0049] Please refer to Figure 1 , Figure 1 A flowchart of the method for detecting financial fraud according to an embodiment of the present application is shown; Figure 1 As shown, the financial fraud detection method includes the following steps S101-S104:
[0050] S101. Obtaining financial data of target users in multiple dimensions; the financial data includes transaction data, user behavior data, and audio and video data; the audio and video data is the audio and video evidence provided by the user during the application and / or transaction process;
[0051] S102, extracting target features of each type of financial data using a feature extraction module that matches the type of financial data, thereby obtaining target features of multiple dimensions;
[0052] S103, fusing the target features of the multiple dimensions to generate a comprehensive feature vector;
[0053] S104: Process the comprehensive feature vector using a pre-trained fraud detection model to detect whether fraudulent behavior exists, and output a fraud detection result.
[0054] The financial fraud detection method is applied to a financial fraud detection system, which can also be referred to as an intelligent decision engine system for identifying financial fraud.
[0055] The financial fraud system mainly includes five parts: data acquisition module, feature extraction module, model building module, decision engine module and result output module.
[0056] In step S101, financial data of multiple dimensions of the target user is obtained; the financial data includes transaction data, user behavior data and audio and video data; the audio and video data is the audio and video evidence provided by the user during the application and / or transaction process.
[0057] Specifically, the data collection module in the financial fraud detection system collects transaction data, user behavior data, and audio and video data from financial institutions.
[0058] The transaction data includes the user's account information and transaction records. The transaction records include transaction time, amount, channel (APP / webpage / offline), etc.; user behavior data includes the user's login behavior, operating habits, etc.; audio and video data includes audio and video evidence provided by the user during the application process.
[0059] Audio and video data includes: audio data and video data.
[0060] Obtain financial data of target users in multiple dimensions. Specifically, when the target user triggers the preset fraud risk rules, the system automatically starts the multi-dimensional data collection process to obtain financial data of target users in multiple dimensions.
[0061] Specifically, the preset fraud risk rules include: high-frequency trading, abnormal login, audio and video verification failure, etc.
[0062] In step S102, target features of each type of financial data are extracted respectively by a feature extraction module that matches the type of financial data, thereby obtaining target features of multiple dimensions.
[0063] The feature extraction module in the financial fraud detection preprocesses and extracts features from the collected data. Specifically, for transaction data and user behavior data, statistical analysis and machine learning algorithms are used to extract key features. For audio and video data, audio and video processing technology and AI algorithms are used to extract audio and video synthesis features, such as audio tampering traces and video frame repetition rate.
[0064] Please refer to Figure 2 The feature extraction module matching the financial data type extracts target features of each type of financial data to obtain target features of multiple dimensions, including the following steps S201-S202:
[0065] S201. Extract target fraud features of the transaction data and user behavior data using statistical analysis algorithms and machine learning algorithms; the target fraud features represent features related to fraudulent behavior;
[0066] S202: For audio and video data, extract audio and video synthesis features as target features of this dimension.
[0067] Specifically, for audio and video data, audio and video synthesis features are extracted, including:
[0068] For audio and video data, extract audio tampering trace features and video frame repetition rate features;
[0069] The audio and video synthesis feature is generated based on the audio tampering trace feature and the video frame repetition rate feature.
[0070] For transaction data and user behavior data, the target fraud features of this dimension are extracted through statistical analysis algorithms and machine learning algorithms. Specifically, data cleaning and standardization are first performed. For example, in some embodiments, missing value processing, data normalization, category coding, time series alignment and other processing are performed.
[0071] Extract the target fraud features of this dimension through statistical analysis algorithms. Specifically, perform univariate statistical analysis and multivariate correlation analysis.
[0072] For example, univariate statistical analysis can calculate the mean, variance, skewness, and kurtosis of transaction amounts, and identify abnormalities in the amount distribution (such as a right-skewed distribution that may indicate large-scale fraudulent transactions); frequency analysis can be performed, such as counting the number of transactions per hour by users, and those exceeding the historical 95% percentile are marked as high-frequency anomalies.
[0073] For example, multivariate correlation analysis can calculate the Spearman correlation between transaction amount and device type and IP geographic location through the correlation coefficient matrix, and discover abnormal correlations (such as sudden large transfers of new devices).
[0074] Based on the analysis results, target fraud features corresponding to transaction data and target fraud features corresponding to user behavior data are constructed.
[0075] In some optional embodiments, for audio and video data, audio tampering trace features and video frame repetition rate features are extracted, including: extracting audio tampering trace features through audio tampering trace detection, and calculating the video frame repetition rate.
[0076] When detecting traces of audio tampering, data preprocessing is performed first. Data preprocessing includes converting the audio into a WAV format with a high sampling rate (such as 44.1kHz) to avoid compression loss; removing background noise, etc.
[0077] When detecting tampering traces on pre-processed audio data, waveform anomaly detection can be performed: analyzing waveform amplitude mutations (such as discontinuous jumps at cuts and splices); and / or spectrum analysis: detecting spectral discontinuities through short-time Fourier transform (STFT) or Mel-spectrogram; and / or phase consistency detection: detecting abnormal fluctuations in the phase spectrum. Synthetic audio may have unnatural phases due to tool processing.
[0078] In some embodiments, detection can also be performed by forging an audio detection model.
[0079] For video data, when calculating the video frame repetition rate, video preprocessing is first performed. Specifically, key frames are extracted at fixed intervals (such as 1 frame per second) or dynamic scene changes (using optical flow method), and all frames are scaled to a uniform resolution (such as 224x224) to reduce computational complexity: RGB is converted to grayscale or YUV channels, and brightness information is separated to reduce noise interference.
[0080] When calculating the video frame repetition rate, it can be achieved through image similarity calculation; specifically, image similarity can be calculated through SSIM (structural similarity), PSNR (peak signal-to-noise ratio), histogram matching, etc., or frame feature vectors can be extracted through a trained CNN (such as VGG, ResNet) to calculate cosine similarity.
[0081] The video frame repetition rate is the ratio of repeated frames to the total number of frames. A high repetition rate indicates a higher probability of being a synthetic video mark.
[0082] In step S103, the target features of the multiple dimensions are fused to generate a comprehensive feature vector.
[0083] Specifically, the target fraud features of the transaction data, the target fraud features of the user behavior data, the audio tampering trace features and the video frame repetition rate features are integrated to generate a comprehensive feature vector.
[0084] In step S104, the comprehensive feature vector is processed by a pre-trained fraud detection model to detect whether fraudulent behavior exists, and a fraud detection result is output.
[0085] Specifically, the pre-trained fraud detection model is constructed and trained based on the model building module of the financial fraud detection system. The model building module is responsible for constructing the fraud detection model. The target sample features of multiple dimensions of the extracted sample data are integrated to form a comprehensive sample feature vector. This comprehensive sample feature vector is then trained using a machine learning algorithm (such as deep learning or support vector machines) to construct a fraud detection model capable of automatically learning and identifying key features of fraudulent behavior.
[0086] The decision engine module is responsible for analyzing and making decisions in real time based on new transaction data, user behavior data, and audio and video data using the built-in fraud detection model. When potential fraud is detected, the decision engine triggers an early warning mechanism and generates a detailed fraud report.
[0087] Please refer to Figure 3 After outputting the fraud detection result, the method further includes the following steps S301-S303:
[0088] S301: When the fraud detection result indicates that fraud exists, extract target attribute information representing attributes of the transaction behavior from the transaction information;
[0089] S302: Process the comprehensive feature vector to generate target fraud evidence information;
[0090] S303: Generate a target fraud report based on a predefined fraud report template, the target attribute information, and the target fraud evidence information.
[0091] The target attribute information includes: user attributes, transaction serial number, timestamp, amount, currency, transaction type, etc.
[0092] The processing of the comprehensive feature vector to generate target fraud evidence information includes at least one of the following:
[0093] Processing the audio and video synthesis features in the comprehensive feature vector to generate audio and video fraud evidence;
[0094] The target fraud features in the comprehensive feature vector are processed to generate quantified behavioral anomaly evidence.
[0095] The audio and video synthesis feature is obtained by extracting audio tampering trace features and video frame repetition rate features from audio and video data.
[0096] The audio and video fraud evidence is the result of analyzing the characteristics of audio tampering traces and video frame repetition rate, such as audio phase discontinuity and video repetition frame rate >30%.
[0097] Behavioral anomaly evidence refers to abnormal behavior obtained based on target fraud feature analysis, such as the same device initiating 10 transactions within 5 minutes.
[0098] Generating a target fraud report based on a predefined fraud report template and the target attribute information and the target fraud evidence information includes: inserting the target attribute information and the target fraud evidence information into the predefined fraud report template to generate the target fraud report.
[0099] For example, the target fraud report is: user AA initiates a transaction to account CC at time BB, wherein the audio phase is discontinuous, and the same device initiates 10 transactions within 5 minutes, which may indicate financial fraud.
[0100] Please refer to Figure 4 After generating the target fraud report, the method further includes the following steps S401-S403:
[0101] S401: When the fraud detection result indicates that fraud exists, generate fraud warning information;
[0102] S402: Acquire target countermeasure information matching the fraudulent behavior detection result from pre-configured countermeasure information;
[0103] S403: Send the target fraud report, fraud warning information, target response information, and fraud behavior detection results to the target terminal device.
[0104] Sending the target fraud report, fraud warning information, target response information, and fraud behavior detection results to the target terminal device includes:
[0105] Based on different pre-configured roles of the target terminal device, the target fraud report and target countermeasure information are processed to obtain the target fraud report and target countermeasure information that match the role; different roles have different permissions;
[0106] The fraud warning information, fraud behavior detection results, target fraud report matching the role, and target response measures information are sent to the target terminal device corresponding to the role.
[0107] Exemplarily, the response measures information includes: automatic interception, manual review, making an early warning call, etc.
[0108] The target countermeasure information is some suggestions for financial fraud. When the roles of the staff are different, the target countermeasure information is different. Therefore, the target countermeasure information is adjusted based on the role information.
[0109] For example, the target response information for the risk department is: Please freeze the account immediately, please contact the security team for processing; the target response information for the security department is: Please freeze the account immediately.
[0110] Different staff members have different permissions to view different content in the targeted fraud reports. For example, a report for a risk control specialist contains all content, while a report for a client representative only contains risk conclusions.
[0111] The target terminal device is the terminal where staff receive warnings and reports, including risk control systems, emails, text messages, internal communication tools, etc.
[0112] The different roles of the target terminal device are specifically the different roles associated with the target terminal device's identifier, such as the role associated with an email address. A role is a predefined set of permissions in the system, bound to responsibilities. The role associated with the target terminal device's identifier (such as email address, mobile phone number, device ID, etc.) is essentially a functional classification of the user's identity in the business scenario, determining the scope of data they can access and the operational permissions they have.
[0113] In some embodiments, processing the target fraud report includes: encrypting part of the content, deleting part of the content, etc.
[0114] In some embodiments, processing the target countermeasure information includes: modifying text in the target countermeasure information, etc.
[0115] In some embodiments, the analysis results and warning information of the decision engine are output to relevant personnel of the financial institution through the result output module in the financial fraud detection system; the output content includes detailed information of the fraudulent behavior, risk assessment results, and recommended response measures, etc., to provide decision support for the financial institution.
[0116] Based on the same inventive concept, the embodiments of the present application also provide a financial fraud detection device corresponding to the financial fraud detection method. Since the principle of solving the problem by the device in the embodiments of the present application is similar to the above-mentioned financial fraud detection method in the embodiments of the present application, the implementation of the device can refer to the implementation of the method, and the repeated parts will not be repeated.
[0117] Please refer to Figure 5 , Figure 5 FIG. 1 shows a schematic diagram of the structure of the financial fraud detection device according to an embodiment of the present application; FIG. Figure 5 As shown, the financial fraud detection device includes:
[0118] Acquisition module 501 is used to acquire financial data of target users in various dimensions; the financial data includes transaction data, user behavior data, and audio and video data; the audio and video data is the audio and video evidence provided by the user during the application and / or transaction process;
[0119] An extraction module 502 is configured to extract target features of each type of financial data using a feature extraction module that matches the type of financial data, thereby obtaining target features of multiple dimensions;
[0120] A fusion module 503 is used to fuse the target features of the multiple dimensions to generate a comprehensive feature vector;
[0121] The detection module 504 is configured to process the comprehensive feature vector using a pre-trained fraud detection model to detect whether fraudulent behavior exists and output a fraudulent behavior detection result.
[0122] In some embodiments, in the financial fraud detection device, the extraction module, when extracting target features of each type of financial data using a feature extraction module matched to the type of financial data to obtain target features of multiple dimensions, is specifically configured to:
[0123] Target fraud features of transaction data and user behavior data are extracted using statistical analysis and machine learning algorithms. These target fraud features represent characteristics related to fraudulent behavior.
[0124] For audio and video data, audio and video synthesis features are extracted as the target features of this dimension.
[0125] In some embodiments, in the financial fraud detection device, the extraction module, when extracting audio and video synthesis features from audio and video data, is specifically configured to:
[0126] For audio and video data, extract audio tampering trace features and video frame repetition rate features;
[0127] The audio and video synthesis feature is generated based on the audio tampering trace feature and the video frame repetition rate feature.
[0128] In some embodiments, the financial fraud detection device further includes:
[0129] a generating module for extracting target attribute information representing attributes of the transaction behavior from the transaction information after outputting the fraud detection result and determining that the fraud detection result indicates the presence of fraud;
[0130] processing the comprehensive feature vector to generate target fraud evidence information;
[0131] A target fraud report is generated based on a predefined fraud report template, the target attribute information, and the target fraud evidence information.
[0132] In some embodiments, in the financial fraud detection device, the generation module, when processing the comprehensive feature vector to generate target fraud evidence information, is specifically configured to perform at least one of the following:
[0133] Processing the audio and video synthesis features in the comprehensive feature vector to generate audio and video fraud evidence;
[0134] The target fraud features in the comprehensive feature vector are processed to generate quantified behavioral anomaly evidence.
[0135] In some embodiments, the financial fraud detection device further includes a sending module for generating fraud warning information when the fraud behavior detection result indicates that fraud behavior exists after generating the target fraud report;
[0136] Obtaining target countermeasure information matching the fraudulent behavior detection result from pre-configured countermeasure information;
[0137] The target fraud report, fraud warning information, target response information and fraud behavior detection results are sent to the target terminal device.
[0138] In some embodiments, the sending module in the financial fraud detection device, when sending the target fraud report, fraud warning information, target response information, and fraud behavior detection result to the target terminal device, is specifically configured to:
[0139] Based on different pre-configured roles of the target terminal device, the target fraud report and target countermeasure information are processed to obtain the target fraud report and target countermeasure information that match the role; different roles have different permissions;
[0140] The fraud warning information, fraud behavior detection results, target fraud report matching the role, and target response measures information are sent to the target terminal device corresponding to the role.
[0141] Based on the same inventive concept, an electronic device corresponding to the financial fraud detection method is also provided in the embodiments of the present application. Since the principle of solving the problem by the electronic device in the embodiments of the present application is similar to the above-mentioned financial fraud detection method in the embodiments of the present application, the implementation of the electronic device can refer to the implementation of the method, and the repeated parts will not be repeated.
[0142] Please refer to Figure 6 , Figure 6 A schematic structural diagram of an electronic device described in an embodiment of the present application is shown; the electronic device includes: a processor, a memory, and a bus, the memory stores machine-readable instructions executable by the processor, and when the electronic device is running, the processor and the memory communicate via the bus, and when the machine-readable instructions are executed by the processor, the steps of the financial fraud detection method are performed.
[0143] Based on the same inventive concept, the embodiments of the present application also provide a computer-readable storage medium corresponding to the financial fraud detection method. Since the principle of solving the problem by the computer-readable storage medium in the embodiments of the present application is similar to the above-mentioned financial fraud detection method in the embodiments of the present application, the implementation of the computer-readable storage medium can refer to the implementation of the method, and the repeated parts will not be repeated.
[0144] A computer-readable storage medium stores a computer program, which executes the steps of the financial fraud detection method when executed by a processor.
[0145] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the system and device described above can refer to the corresponding process in the method embodiment, and will not be repeated in this application. In the several embodiments provided in this application, it should be understood that the disclosed system, device and method can be implemented in other ways. The device embodiments described above are merely schematic. For example, the division of the modules is only a logical function division. There may be other division methods in actual implementation. For example, multiple modules or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some communication interfaces, indirect coupling or communication connection of devices or modules, which can be electrical, mechanical or other forms.
[0146] The modules described as separate components may or may not be physically separate, and the components shown as modules may or may not be physical units, that is, they may be located in one place or distributed across multiple network elements. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.
[0147] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.
[0148] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a non-volatile computer-readable storage medium that is executable by a processor. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, platform server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes various media that can store program codes, such as a USB flash drive, a mobile hard disk, a ROM, a RAM, a magnetic disk, or an optical disk.
[0149] The above are only specific embodiments of the present application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.
Claims
1. A financial fraud detection method, characterized in that: The method comprises the following steps: Obtaining financial data of target users in multiple dimensions; the financial data includes transaction data, user behavior data, and audio and video data; the audio and video data is the audio and video evidence provided by the user during the application and / or transaction process; Through the feature extraction module that matches the financial data type, the target features of each type of financial data are extracted respectively, and the target features of multiple dimensions are obtained; Fusing the target features of the multiple dimensions to generate a comprehensive feature vector; The comprehensive feature vector is processed by a pre-trained fraud detection model to detect whether fraudulent behavior exists and output a fraud detection result.
2. The financial fraud detection method according to claim 1, characterized in that: The target features of each type of financial data are extracted by a feature extraction module that matches the type of financial data, and target features of multiple dimensions are obtained, including: Target fraud features of transaction data and user behavior data are extracted using statistical analysis and machine learning algorithms. These target fraud features represent characteristics related to fraudulent behavior. For audio and video data, audio and video synthesis features are extracted as the target features of this dimension.
3. The financial fraud detection method according to claim 2, characterized in that: For audio and video data, extract audio and video synthesis features, including: For audio and video data, extract audio tampering trace features and video frame repetition rate features; The audio and video synthesis feature is generated based on the audio tampering trace feature and the video frame repetition rate feature.
4. The financial fraud detection method according to claim 1, characterized in that: After outputting the fraud detection result, the method further includes: When the fraud detection result indicates that fraud exists, extracting target attribute information representing attributes of the transaction behavior from the transaction information; processing the comprehensive feature vector to generate target fraud evidence information; A target fraud report is generated based on a predefined fraud report template, the target attribute information, and the target fraud evidence information.
5. The financial fraud detection method according to claim 4, characterized in that: Processing the comprehensive feature vector to generate target fraud evidence information includes at least one of the following: Processing the audio and video synthesis features in the comprehensive feature vector to generate audio and video fraud evidence; The target fraud features in the comprehensive feature vector are processed to generate quantified behavioral anomaly evidence.
6. The financial fraud detection method according to claim 4, characterized in that: After generating the targeted fraud report, the method further includes: When the fraud detection result indicates that fraud exists, generating fraud warning information; Obtaining target countermeasure information matching the fraudulent behavior detection result from pre-configured countermeasure information; The target fraud report, fraud warning information, target response information and fraud behavior detection results are sent to the target terminal device.
7. The financial fraud detection method according to claim 6, characterized in that: Sending the target fraud report, fraud warning information, target response information, and fraud behavior detection results to the target terminal device includes: Based on different pre-configured roles of the target terminal device, the target fraud report and target countermeasure information are processed to obtain the target fraud report and target countermeasure information that match the role; different roles have different permissions; The fraud warning information, fraud behavior detection results, target fraud report matching the role, and target response measures information are sent to the target terminal device corresponding to the role.
8. A financial fraud detection device, characterized in that: The device comprises: An acquisition module is used to obtain financial data of target users in multiple dimensions; the financial data includes transaction data, user behavior data, and audio and video data; the audio and video data is the audio and video evidence provided by the user during the application and / or transaction process; An extraction module is used to extract target features of each type of financial data through a feature extraction module that matches the type of financial data, thereby obtaining target features of multiple dimensions; A fusion module, configured to fuse the target features of the multiple dimensions to generate a comprehensive feature vector; The detection module is used to process the comprehensive feature vector through a pre-trained fraud detection model to detect whether fraudulent behavior exists and output a fraud detection result.
9. An electronic device, characterized in that: include: A processor, a memory, and a bus, wherein the memory stores machine-readable instructions executable by the processor. When the electronic device is running, the processor and the memory communicate via the bus. When the machine-readable instructions are executed by the processor, the steps of the financial fraud detection method according to any one of claims 1 to 7 are performed.
10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, which, when executed by a processor, executes the steps of the financial fraud detection method according to any one of claims 1 to 7.