Risk level determination methods, apparatus, equipment, media and procedures products

Abstract

This application provides a method for determining risk levels. It can be applied to the fields of big data technology and artificial intelligence technology, involving the application of large-scale models in information security and fintech. The method includes: acquiring target data; extracting semantic features from the target data to generate target feature semantic labels; inputting the target feature semantic labels into a pre-trained first risk model to generate a first risk label and a first risk value; inputting the first risk label into a pre-trained second risk model to generate a second risk value; generating a target risk value for the target data based on the first risk value and the second risk value; and determining the target risk level of the target data based on the target risk value. This application also provides a risk level determination apparatus, device, storage medium, and program product.

Description

Technical Field

[0001] This application relates to the fields of big data technology and artificial intelligence technology, specifically to the application of large models in information security and financial technology, and particularly to a risk level determination method, apparatus, equipment, medium and program product. Background Technology

[0002] In high-security environments such as financial trading rooms, unusual behaviors, including the unauthorized use of mobile phones, pose a serious security threat. In particular, using mobile phones to take photos, record videos, or transmit sensitive transaction information can lead to actions that severely violate the principles of fairness, impartiality, and transparency in the financial market. These actions not only directly threaten the company's trade secrets and customer assets but also seriously disrupt the normal financial market order, harm the legitimate rights and interests of investors, and may trigger severe penalties from regulatory agencies.

[0003] Traditional mobile phone detection methods typically rely on physical shielding, signal interference, or manual monitoring. These methods have limitations, failing to distinguish between legitimate use and illegal activities, or struggling to accurately identify subtle violation patterns in complex and dynamic environments in real time. Existing automated monitoring systems often focus on coarse-grained detection, such as determining the presence or activation of a mobile phone, but struggle to further identify the specific illegal uses of the phone, such as taking photos, recording videos, transmitting sensitive data, or accessing unauthorized applications. This deficiency in coarse-grained detection leads to numerous false alarms and may also miss truly serious, more granular anomalies. This results in low accuracy and a high false alarm rate in risk level identification corresponding to granularity, while also wasting computational resources and reducing computational efficiency, posing significant challenges to the security management of financial trading rooms. Summary of the Invention

[0004] In view of the above problems, this application provides a method, apparatus, equipment, medium and procedure for determining risk level.

[0005] According to a first aspect of this application, a method for determining a risk level is provided. The method includes: acquiring target data; extracting semantic features from the target data to generate target feature semantic labels; inputting the target feature semantic labels into a pre-trained first risk model to generate a first risk label and a first risk value, wherein the first risk model is trained based on first historical data, and the first historical data has a first risk level; inputting the first risk label into a pre-trained second risk model to generate a second risk value, wherein the second risk model is trained based on second historical data, and the second historical data has a second risk level, the second risk level being lower than the first risk level; generating a target risk value for the target data based on the first risk value and the second risk value; and determining a target risk level for the target data based on the target risk value.

[0006] According to an embodiment of this application, the target data includes target video data. Acquiring the target data and extracting semantic features from the target data to generate target feature semantic tags includes: acquiring the target video data; performing frame extraction processing on the target video data to generate target image data; performing text conversion on the target image data to generate target text data; and extracting semantic features from the target text data to generate the target feature semantic tags.

[0007] According to an embodiment of this application, text conversion of the target image data to generate target text data includes: extracting abnormal features from the target image data using a basic encoder to generate abnormal feature data; and performing text processing on the abnormal feature data to generate the target text data.

[0008] According to an embodiment of this application, generating a target risk value for the target data based on the first risk value and the second risk value includes: presetting a first weight and a second weight; calculating the product of the first risk value and the first weight to generate a first association value; calculating the product of the second risk value and the second weight to generate a second association value; and generating the target risk value based on the first association value and the second association value.

[0009] According to an embodiment of this application, presetting a first weight and a second weight includes: calculating the difference between the first risk level and the second risk level; and presetting the first weight and the second weight based on the difference between the first risk level and the second risk level, wherein the difference between the first weight and the second weight is mapped to the difference between the first risk level and the second risk level.

[0010] According to an embodiment of this application, the method further includes: calculating a common loss value for the first risk model and the second risk model based on the first risk level, the second risk level, and the difference between the first risk level and the second risk level; and updating the first risk model and the second risk model based on the common loss value.

[0011] According to an embodiment of this application, the method further includes: detecting the first risk value using the second risk value; if the detection fails, performing a rollback and sending a notification.

[0012] According to a second aspect of this application, a risk level determination apparatus is provided, comprising: a first acquisition module for acquiring target data, extracting semantic features from the target data, and generating target feature semantic labels; a first generation module for inputting the target feature semantic labels into a pre-trained first risk model to generate a first risk label and a first risk value, wherein the first risk model is trained based on first historical data, and the first historical data has a first risk level; a second generation module for inputting the first risk label into a pre-trained second risk model to generate a second risk value, wherein the second risk model is trained based on second historical data, and the second historical data has a second risk level, the second risk level being lower than the first risk level; a third generation module for generating a target risk value for the target data based on the first risk value and the second risk value; and a first determination module for determining the target risk level of the target data based on the target risk value.

[0013] According to a third aspect of this application, an electronic device is provided, comprising: one or more processors; and a memory for storing one or more computer programs, wherein the one or more processors execute the one or more computer programs to implement the steps of the method described above.

[0014] A fourth aspect of this application also provides a computer-readable storage medium having a computer program or instructions stored thereon, which, when executed by a processor, implement the steps of the above-described method.

[0015] The fifth aspect of this application also provides a computer program product, including a computer program or instructions that, when executed by a processor, implement the steps of the above-described method.

[0016] This application employs a two-layer risk model collaborative mechanism. Data is input into a model trained at a high-risk level, which outputs a first risk label and a first risk score. The first risk label is then input into a model trained at a lower risk level, which outputs a second risk score. This precise risk level determination significantly improves the accuracy of risk detection and the reliability of risk level determination, reduces the false alarm rate, and achieves the technical effects of saving computer resources and improving computational efficiency. It solves the technical problems of low accuracy and high false alarm rate in existing risk level identification technologies, which lead to wasted computational resources and low computational efficiency. Attached Figure Description

[0017] The above-mentioned contents, other objects, features and advantages of this application will become clearer from the following description of embodiments with reference to the accompanying drawings, in which:

[0018] Figure 1 The illustrations depict application scenarios of risk level determination methods, apparatus, devices, media, and program products according to embodiments of this application.

[0019] Figure 2 A flowchart illustrating a risk level determination method according to an embodiment of this application is shown schematically.

[0020] Figure 3 This illustration schematically shows a flowchart of generating target feature semantic labels in a risk level determination method according to an embodiment of this application;

[0021] Figure 4 This illustration schematically shows a flowchart of the risk level determination method according to an embodiment of the present application, in which target text data is generated;

[0022] Figure 5 This illustration schematically shows a flowchart of the method for determining a risk level according to an embodiment of the present application, in which a target risk value is generated;

[0023] Figure 6 A flowchart illustrating the preset weights in the risk level determination method according to an embodiment of this application is shown.

[0024] Figure 7 This schematic diagram illustrates a structural block diagram of a risk level determination device according to an embodiment of the present application;

[0025] Figure 8 A block diagram schematically illustrates an electronic device suitable for implementing a risk level determination method according to an embodiment of this application. Detailed Implementation

[0026] The embodiments of this application will now be described with reference to the accompanying drawings. However, it should be understood that these descriptions are exemplary only and are not intended to limit the scope of this application. In the following detailed description, numerous specific details are set forth to provide a thorough understanding of the embodiments of this application for ease of explanation. However, it will be apparent that one or more embodiments may be implemented without these specific details. Furthermore, descriptions of well-known structures and technologies are omitted in the following description to avoid unnecessarily obscuring the concepts of this application.

[0027] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of this application. The terms “comprising,” “including,” etc., as used herein indicate the presence of the stated features, steps, operations, and / or components, but do not exclude the presence or addition of one or more other features, steps, operations, or components.

[0028] All terms used herein (including technical and scientific terms) have the meanings commonly understood by those skilled in the art, unless otherwise defined. It should be noted that the terms used herein are to be interpreted in a manner consistent with the context of this specification, and not in an idealized or overly rigid way.

[0029] When using expressions such as "at least one of A, B and C", they should generally be interpreted in accordance with the meaning that is commonly understood by those skilled in the art (e.g., "a system having at least one of A, B and C" should include, but is not limited to, a system having A alone, a system having B alone, a system having C alone, a system having A and B, a system having A and C, a system having B and C, and / or a system having A, B and C, etc.).

[0030] The accompanying drawings show some block diagrams and / or flowcharts. It should be understood that some blocks or combinations thereof in the block diagrams and / or flowcharts can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable control device, so that when executed by the processor, these instructions can create means for implementing the functions / operations described in these block diagrams and / or flowcharts.

[0031] First, the technical terms used in this article are explained as follows:

[0032] The basic encoder is a model based on a neural network model architecture that has been pre-trained on massive amounts of data.

[0033] A classifier is an algorithm or model that automatically predicts the category of a sample based on input data features.

[0034] This application provides a risk level determination method, comprising: acquiring target data; extracting semantic features from the target data to generate target feature semantic labels; inputting the target feature semantic labels into a pre-trained first risk model to generate a first risk label and a first risk value, wherein the first risk model is trained based on first historical data, and the first historical data has a first risk level; inputting the first risk label into a pre-trained second risk model to generate a second risk value, wherein the second risk model is trained based on second historical data, and the second historical data has a second risk level, the second risk level being lower than the first risk level; generating a target risk value for the target data based on the first risk value and the second risk value; and determining a target risk level for the target data based on the target risk value.

[0035] According to the embodiments of this application, by constructing a two-layer risk model collaborative mechanism, data is input into a model trained at a high-risk level to output a first risk label and a first risk score. Then, the first risk label is input into a model trained at a lower risk level to output a second risk score. This technical approach, which ultimately achieves accurate risk level determination, significantly improves the accuracy of risk detection and the reliability of risk level determination, reduces the false alarm rate, and achieves the technical effects of saving computer resources and improving computational efficiency. It solves the technical problems of low accuracy and high false alarm rate in risk level identification in existing technologies, which lead to wasted computational resources and low computational efficiency.

[0036] Figure 1 The illustration schematically depicts application scenarios of the risk level determination method, apparatus, device, medium, and program product according to embodiments of this application. It should be noted that... Figure 1 The examples shown are merely examples of scenarios in which the embodiments of this application can be applied, to help those skilled in the art understand the technical content of this application, but do not mean that the embodiments of this application cannot be used in other devices, systems, environments or scenarios.

[0037] like Figure 1 As shown, application scenario 100 according to this embodiment may include application scenarios with determined risk levels. Network 104 serves as a medium for providing communication links between the first terminal device 101, the second terminal device 102, the third terminal device 103, and the server 105. Network 104 may include various connection types, such as wired or wireless communication links or fiber optic cables, etc.

[0038] Users can use the first terminal device 101, the second terminal device 102, and the third terminal device 103 to interact with the server 105 via the network 104 to receive or send messages, etc. Various communication client applications can be installed on the first terminal device 101, the second terminal device 102, and the third terminal device 103, such as shopping applications, web browser applications, search applications, instant messaging tools, email clients, social media platform software, etc. (for example only).

[0039] The first terminal device 101, the second terminal device 102, and the third terminal device 103 can be various electronic devices with displays and support web browsing, including but not limited to smartphones, tablets, laptops, and desktop computers.

[0040] Server 105 can be a server that provides various services, such as a backend management server that supports websites browsed by users using the first terminal device 101, the second terminal device 102, and the third terminal device 103 (this is just an example). The backend management server can analyze and process data such as received user requests, and feed back the processing results (such as web pages, information, or data obtained or generated according to user requests) to the terminal devices.

[0041] It should be noted that the risk level determination method provided in this application embodiment can generally be executed by server 105. Correspondingly, the risk level determination device provided in this application embodiment can generally be located in server 105. The risk level determination method provided in this application embodiment can also be executed by a server or server cluster that is different from server 105 and capable of communicating with the first terminal device 101, the second terminal device 102, the third terminal device 103, and / or server 105. Correspondingly, the risk level determination device provided in this application embodiment can also be located in a server or server cluster that is different from server 105 and capable of communicating with the first terminal device 101, the second terminal device 102, the third terminal device 103, and / or server 105.

[0042] It should be understood that Figure 1 The number of terminal devices, networks, and servers shown is merely illustrative. Depending on implementation needs, any number of terminal devices, networks, and servers can be included.

[0043] The following will be based on Figure 1 The described scene, through Figures 2-6 The risk level determination method of the disclosed embodiments is described in detail. It should be noted that the above application scenarios are shown only to facilitate understanding of the spirit and principles of this application, and the implementation methods of this application are not limited in any way. On the contrary, the implementation methods of this application can be applied to any applicable scenario.

[0044] Figure 2 A flowchart illustrating a risk level determination method according to an embodiment of this application is shown schematically.

[0045] like Figure 2 As shown, the method 200 includes steps S201 to S205.

[0046] Step S201: Obtain target data, extract semantic features from the target data, and generate target feature semantic labels.

[0047] For example, target data can include multimodal data such as video data, image data, and text data. In high-security environments such as financial trading rooms, camera video streams can be acquired as target video data for semantic feature extraction.

[0048] In the technical solution of this application, the user information (including but not limited to user personal information, user image information, user device information, such as location information) and data (including but not limited to data used for analysis, stored data, and displayed data) involved are all information and data authorized by the user or fully authorized by all parties. Furthermore, the collection, storage, use, processing, transmission, provision, disclosure, and application of related data all comply with relevant laws, regulations, and standards, take necessary confidentiality measures, do not violate public order and good morals, and provide corresponding operation entry points for users to choose to authorize or refuse.

[0049] Figure 3 The flowchart illustrating the generation of target feature semantic labels in the risk level determination method according to an embodiment of this application is shown.

[0050] like Figure 3 As shown, the method 300 includes steps S301 to S303.

[0051] Step S301: Obtain target video data, perform frame extraction processing on the target video data, and generate target image data.

[0052] For example, frame extraction processing of target video data includes: acquiring target video information, pre-setting a frame extraction time interval based on the target video information, and performing frame extraction processing on the target video data according to the frame extraction time interval. Frame extraction can also be based on changes in video content, specifically including frame extraction based on detecting changes in motion within the video frame. Alternatively, frame extraction can be performed using key object recognition. Specifically, in high-security environments such as financial trading rooms, a mobile phone can be designated as a key object, and video stream data from cameras in the financial trading room can be collected for recognition and frame extraction to generate an image set as target image data.

[0053] Step S302: Perform text conversion on the target image data to generate target text data.

[0054] Figure 4 The flowchart illustrating the generation of target text data in the risk level determination method according to an embodiment of this application is shown in the illustration.

[0055] like Figure 4 As shown, the method 400 includes steps S401 to S402.

[0056] Step S401: Extract abnormal features from the target image data using the basic encoder to generate abnormal feature data.

[0057] For example, anomalous features can include features related to mobile phone use, which may include, but are not limited to, human posture features (e.g., operating the screen, making a call, taking a picture of the screen), and the phone's position and status (e.g., whether it is in the hand or facing the screen). An anomalous feature data can be generated by extracting these mobile phone use-related features from image data using an encoder.

[0058] Step S402: Perform text processing on the abnormal feature data to generate the target text data.

[0059] For example, image data with features related to mobile phone use can be input into a pre-trained computer vision model to generate natural language descriptions of the image features as target text data.

[0060] By employing techniques such as targeted extraction of abnormal features and image textification, data dimensionality reduction and precise data positioning can be achieved, significantly improving the reliability of the generated target text data.

[0061] Return to reference Figure 3 In step S303, semantic features are extracted from the target text data to generate the target feature semantic tags.

[0062] For example, semantic features can be extracted from the target text data describing the natural language features of an image using natural language processing techniques to generate semantic labels for the target features.

[0063] By constructing a transformation process from video data to image data, from image data to text data, and then from text data to video data and finally to target feature semantic labels, the reliability of generating target feature semantic labels is significantly improved. These target feature semantic labels, as input data for the subsequent first-risk model, are equivalent to precisely refining the model's input data, thereby improving the accuracy of subsequent model outputs and the model's recognition efficiency.

[0064] Return to reference Figure 2In step S202, the target feature semantic label is input into a pre-trained first risk model to generate a first risk label and a first risk value, wherein the first risk model is trained and generated based on first historical data, and the first historical data has a first risk level.

[0065] For example, the risk level of a target feature semantic label can be correlated with the granularity of the abnormal historical data used to train the risk model. The finer the granularity of the abnormal historical data, the higher the risk level of the corresponding abnormal historical data. The first risk model can use a classifier, which can be the finest-grained first classifier trained based on the first historical data. The goal of the finest-grained classifier is to accurately identify the finest-grained and most risky abnormal mobile phone behavior, such as whether the mobile phone is recording video or accessing sensitive information, and whether it involves the leakage of transaction information.

[0066] For example, abnormal historical data indicating the phone's presence within a restricted trading area represents the coarsest level of abnormal historical data, corresponding to the lowest risk level (Level 1). Abnormal historical data indicating the phone is visible and in the hands of an operator represents a relatively coarse level of abnormal historical data, corresponding to a lower risk level (Level 2). Abnormal historical data indicating the phone is being actively used (e.g., screen operation or phone call) represents a relatively fine level of abnormal historical data, corresponding to a higher risk level (Level 3). Abnormal historical data indicating the phone is recording / taking photos, accessing disabled applications, or transmitting data over external networks—activities that may involve the leakage of transaction information—represents the finest level of abnormal historical data, corresponding to the highest risk level (Level 4).

[0067] The aforementioned abnormal historical data are merely examples. The abnormal historical data in this application includes, but is not limited to: a mobile phone appearing in a restricted area, a mobile phone being visible and in the hands of an operator, a mobile phone being actively used, a mobile phone recording the screen or taking photos, a mobile phone accessing a disabled application, or a mobile phone transmitting data. The abnormal historical data in this application includes all mobile phone-related transaction information leaks.

[0068] Step S203: Input the first risk label into the pre-trained second risk model to generate a second risk value, wherein the second risk model is trained and generated based on second historical data, the second historical data has a second risk level, and the second risk level is less than the first risk level.

[0069] For example, the second risk model can also output a second risk label. This second risk model can also use a classifier, which could be a finer-grained second classifier trained based on second historical data. Furthermore, multiple classifiers, such as a third or fourth classifier, can be chained between the first and second classifiers, ensuring that the training data for all classifiers from the first to the second classifier are arranged in descending order of risk level. The output of the finer-grained classifier serves as the input to a coarser-grained classifier, allowing the classifier to prioritize and capture features related to the most granular abnormal behavior during the learning process. The learning of fine-grained features is given higher priority, resulting in more discriminative features.

[0070] In trading room environments, even abnormal mobile phone behaviors at the same coarse-grained level can have varying potential risks and severity. For example, at the granularity of "the phone is being actively used," the harm levels of "viewing time" and "screen capture" are drastically different, with the latter potentially directly linked to serious abnormal behavior. Traditional metrics based on the nearest common parent node struggle to accurately reflect these subtle differences. This paper introduces a learnable intra-granularity anomaly discrepancy matrix, designed to adaptively capture subtle semantic or severity differences between different categories of abnormal behavior within the same granularity. The diagonal elements of the matrix are masked to zero. An adaptive intra-granularity discrepancy learning module is introduced to learn a learnable intra-granularity anomaly discrepancy matrix to capture subtle differences or severity among different abnormal behaviors within the same granularity.

[0071] The first risk value is tested using the second risk value. If the test fails, a rollback is performed and a notification is sent.

[0072] For example, if the first risk value is high-risk, the second risk value must also be high-risk; if the first risk value is low-risk, the second risk value could be either high-risk or low-risk. If the test result shows that the first risk value is high-risk and the second risk value is low-risk, it indicates that the first risk model classification is incorrect, and the data should be rolled back and a notification sent to relevant personnel.

[0073] By detecting the first risk value using the second risk value, the second risk value can supervise the first risk value and roll back in a timely manner, thereby improving the reliability of the first risk value.

[0074] The aforementioned detection method can also be applied to scenarios where multiple classifiers are cascaded, allowing each classifier to be updated not only based on its own level of anomaly detection risk but also supervised by its coarser-grained anomaly detection tasks. This approach promotes interaction between classifiers of different granularities, effectively utilizing hierarchical anomaly information, thereby alleviating the "granularity competition" problem and ensuring that fine-grained anomaly features are fully learned and represented.

[0075] Step S204: Based on the first risk value and the second risk value, generate the target risk value of the target data.

[0076] Figure 5 The flowchart illustrating the generation of a target risk value in a risk level determination method according to an embodiment of this application is shown in the illustration.

[0077] like Figure 5 As shown, the method 500 includes steps S501 to S504.

[0078] Step S501: Preset the first weight and the second weight.

[0079] Figure 6 The flowchart illustrating the preset weights in the risk level determination method according to an embodiment of this application is shown in the illustration.

[0080] like Figure 6 As shown, the method 600 includes steps S601 to S602.

[0081] Step S601: Calculate the difference between the first risk level and the second risk level.

[0082] Step S602: Based on the difference between the first risk level and the second risk level, preset the first weight and the second weight, wherein the difference between the first weight and the second weight is mapped to the difference between the first risk level and the second risk level.

[0083] For example, the difference between the first weight and the second weight increases as the difference between the first risk level and the second risk level increases.

[0084] By presetting a first weight and a second weight based on the difference between the first risk level and the second risk level, the difference between the first weight and the second weight changes with the difference between the first risk level and the second risk level. This ensures the reliability of the preset first weight and second weight and further improves the accuracy of the target risk value.

[0085] Return to reference Figure 5 In step S502, the product of the first risk value and the first weight is calculated to generate the first association value.

[0086] Step S503: Calculate the product of the second risk value and the second weight to generate the second association value.

[0087] Step S504: Generate the target risk value based on the first association value and the second association value.

[0088] By assigning weights to the first and second risk values ​​when calculating the target risk value, the accuracy and precision of generating the target risk value are improved.

[0089] Return to reference Figure 2 In step S205, the target risk level of the target data is determined based on the target risk value.

[0090] For example, by using a target risk value, the target risk level of the target data can be determined. Detected abnormal mobile phone behavior can be analyzed and alerts of the corresponding level can be triggered.

[0091] According to embodiments of this application, a common loss value for the first risk model and the second risk model can be calculated based on the first risk level, the second risk level, and the difference between the first risk level and the second risk level; and the first risk model and the second risk model can be updated based on the common loss value.

[0092] For example, the formula for the common loss value is:

[0093]

[0094] in, For the common loss value, It is the weight at each granularity level. Based on parameters preset according to the business scenario, h represents the risk level of the training history data corresponding to the current model, and H represents the maximum value of the risk level. The loss value is the difference in risk levels between the two risk models. This represents the loss value for risk level h.

[0095] In the application scenario of the trading room, by adjusting Parameters can control the contribution of different granular levels to the overall loss. In trading room scenarios, more granular and dangerous abnormal behaviors should be given higher weights to ensure that the system prioritizes learning and detecting these key risks.

[0096] By calculating the common loss value of the first risk model and the second risk model using the first risk level, the second risk level, and the difference between the first risk level and the second risk level, and then updating the first risk model and the second risk model, iterative updates of the first risk model and the second risk model can be achieved, thereby improving the accuracy of subsequent data in judging the target risk level.

[0097] According to the embodiments of this application, the granularity competition problem can be effectively solved, enabling the model to fully learn and distinguish various subtle mobile phone violations within the trading room. It has a stronger ability to identify high-risk behaviors, especially those related to serious anomalies such as transaction information leakage, thus improving the accuracy of fine-grained anomaly detection. Adaptive intra-granularity difference learning allows the system to identify the severity differences of different anomalies within the same granularity. Even if false alarms occur, it tends to classify them into anomalies of similar severity, thereby reducing serious false alarms and more effectively preventing and combating financial violations, effectively lowering the false alarm rate and updating the model. Based on the granularity level of the detected anomalies, risk alerts of varying severity, from "visible on mobile phone" to "mobile phone usage" and "unauthorized filming," can be provided to help security personnel conduct tiered responses, thereby more effectively preventing and combating financial violations and providing multi-level risk warnings.

[0098] Figure 7 A schematic block diagram of a risk level determination device according to an embodiment of this application is shown.

[0099] like Figure 7 As shown, the device 700 includes: a first acquisition module 701, a first generation module 702, a second generation module 703, a third generation module 704, and a first determination module 705.

[0100] The first acquisition module 701 is used to acquire target data, extract semantic features from the target data, and generate target feature semantic labels. In one embodiment, the first acquisition module 701 can be used to execute step S301 described above.

[0101] The first acquisition module 701 includes: a second acquisition module, a fourth generation module, and a fifth generation module.

[0102] The second acquisition module is used to acquire target video data, perform frame extraction processing on the target video data, and generate target image data. In one embodiment, the second acquisition module can be used to execute step S301 described above, which will not be repeated here.

[0103] The fourth generation module is used to perform text conversion on the target image data to generate target text data. In one embodiment, the fourth generation module can be used to perform step S302 described above.

[0104] The fourth generation module includes the sixth generation module and the seventh generation module.

[0105] The sixth generation module is used to extract abnormal features from the target image data using the basic encoder, and generate abnormal feature data. In one embodiment, the sixth generation module can be used to perform step S401 described above, which will not be repeated here.

[0106] The seventh generation module is used to perform text processing on the abnormal feature data to generate the target text data. In one embodiment, the seventh generation module can be used to execute step S402 described above, which will not be repeated here.

[0107] The fifth generation module is used to extract semantic features from the target text data and generate the target feature semantic tags. In one embodiment, the fifth generation module can be used to perform step S303 described above, which will not be repeated here.

[0108] The first generation module 702 is used to input the target feature semantic label into a pre-trained first risk model to generate a first risk label and a first risk value, wherein the first risk model is trained and generated based on first historical data, and the first historical data has a first risk level. In one embodiment, the first generation module 702 can be used to execute step S202 described above, which will not be repeated here.

[0109] The second generation module 703 is used to input the first risk label into a pre-trained second risk model to generate a second risk value. The second risk model is trained based on second historical data, which has a second risk level, and this second risk level is lower than the first risk level. In one embodiment, the second generation module 703 can be used to execute step S203 as described above, which will not be repeated here.

[0110] The third generation module 704 is used to generate a target risk value for the target data based on the first risk value and the second risk value. In one embodiment, the third generation module 704 can be used to perform step S204 described above.

[0111] The third generation module 704 includes: a first preset module, a first calculation module, a second calculation module, and a third calculation module.

[0112] The first preset module is used to preset a first weight and a second weight. In one embodiment, the first preset module can be used to execute step S501 described above.

[0113] The first preset module includes: the fourth calculation module and the second preset module.

[0114] The fourth calculation module is used to calculate the difference between the first risk level and the second risk level. In one embodiment, the fourth calculation module can be used to perform step S601 described above, which will not be repeated here.

[0115] The second preset module is used to preset the first weight and the second weight based on the difference between the first risk level and the second risk level, wherein the difference between the first weight and the second weight is mapped to the difference between the first risk level and the second risk level. In one embodiment, the second preset module can be used to execute step S602 described above, which will not be repeated here.

[0116] The first calculation module is used to calculate the product of the first risk value and the first weight to generate a first association value. In one embodiment, the first calculation module can be used to execute step S502 described above, which will not be repeated here.

[0117] The second calculation module is used to calculate the product of the second risk value and the second weight to generate a second correlation value. In one embodiment, the second calculation module can be used to perform step S503 described above, which will not be repeated here.

[0118] The third calculation module is used to generate the target risk value based on the first association value and the second association value. In one embodiment, the third calculation module can be used to perform step S504 described above, which will not be repeated here.

[0119] The first determining module 705 is used to determine the target risk level of the target data based on the target risk value. In one embodiment, the first determining module 705 can be used to execute step S205 described above, which will not be repeated here.

[0120] In addition, the device also includes a first update module and a first detection module.

[0121] The first update module is used to calculate the common loss value of the first risk model and the second risk model based on the first risk level, the second risk level, and the difference between the first risk level and the second risk level; and to update the first risk model and the second risk model based on the common loss value.

[0122] The first detection module is used to detect the first risk value using the second risk value. If the detection fails, a rollback is performed and a notification is sent.

[0123] According to embodiments of this application, any multiple modules among the first acquisition module 701, first generation module 702, second generation module 703, third generation module 704, and first determination module 705 can be combined into one module, or any one of these modules can be split into multiple modules. Alternatively, at least some of the functions of one or more of these modules can be combined with at least some of the functions of other modules and implemented in one module. According to embodiments of this application, at least one of the first acquisition module 701, first generation module 702, second generation module 703, third generation module 704, and first determination module 705 can be at least partially implemented as hardware circuitry, such as a field-programmable gate array (FPGA), a programmable logic array (PLA), a system-on-a-chip, a system-on-a-substrate, a system-on-package, an application-specific integrated circuit (ASIC), or implemented in hardware or firmware by any other reasonable means of integrating or packaging the circuitry, or implemented in any one of the three implementation methods of software, hardware, and firmware, or in a suitable combination of any of these. Alternatively, at least one of the first acquisition module 701, the first generation module 702, the second generation module 703, the third generation module 704, and the first determination module 705 can be implemented at least partially as a computer program module, which can perform corresponding functions when the computer program module is run.

[0124] Figure 8 A block diagram schematically illustrates an electronic device suitable for implementing a risk level determination method according to an embodiment of this application.

[0125] like Figure 8 As shown, an electronic device 800 according to an embodiment of this application includes a processor 801, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 802 or a program loaded from a storage portion 808 into a random access memory (RAM) 803. The processor 801 may include, for example, a general-purpose microprocessor (e.g., a CPU), an instruction set processor and / or an associated chipset and / or a special-purpose microprocessor (e.g., an application-specific integrated circuit (ASIC)), etc. The processor 801 may also include onboard memory for caching purposes. The processor 801 may include a single processing unit or multiple processing units for performing different actions of the method flow according to an embodiment of this application.

[0126] RAM 803 stores various programs and data required for the operation of electronic device 800. Processor 801, ROM 802, and RAM 803 are interconnected via bus 804. Processor 801 executes various operations of the method flow according to embodiments of this application by executing programs in ROM 802 and / or RAM 803. It should be noted that the programs may also be stored in one or more memories other than ROM 802 and RAM 803. Processor 801 may also execute various operations of the method flow according to embodiments of this application by executing programs stored in said one or more memories.

[0127] According to embodiments of this application, the electronic device 800 may further include an input / output (I / O) interface 805, which is also connected to a bus 804. The electronic device 800 may also include one or more of the following components connected to the input / output (I / O) interface 805: an input section 806 including a keyboard, mouse, etc.; an output section 807 including a cathode ray tube (CRT), liquid crystal display (LCD), etc., and a speaker, etc.; a storage section 808 including a hard disk, etc.; and a communication section 809 including a network interface card such as a LAN card, modem, etc. The communication section 809 performs communication processing via a network such as the Internet. A drive 810 is also connected to the input / output (I / O) interface 805 as needed. A removable medium 811, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., is installed on the drive 810 as needed so that computer programs read from it can be installed into the storage section 808 as needed.

[0128] This application also provides a computer-readable storage medium, which may be included in the device / apparatus / system described in the above embodiments; or it may exist independently and not assembled into the device / apparatus / system. The computer-readable storage medium carries one or more programs, which, when executed, implement the method according to the embodiments of this application.

[0129] According to embodiments of this application, the computer-readable storage medium can be a non-volatile computer-readable storage medium, such as including but not limited to: portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this application, the computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. For example, according to embodiments of this application, the computer-readable storage medium may include ROM 802 and / or RAM 803 and / or one or more memories other than ROM 802 and RAM 803 described above.

[0130] Embodiments of this application also include a computer program product comprising a computer program containing program code for performing the methods shown in the flowchart. When the computer program product is run on a computer system, the program code is used to enable the computer system to implement the risk level determination method provided in the embodiments of this application.

[0131] When the computer program is executed by the processor 801, it performs the functions defined in the system / apparatus of this application embodiment. According to the embodiments of this application, the systems, apparatuses, modules, units, etc., described above can be implemented by computer program modules.

[0132] In one embodiment, the computer program may rely on a tangible storage medium such as an optical storage device or a magnetic storage device. In another embodiment, the computer program may also be transmitted and distributed in the form of signals over a network medium, and may be downloaded and installed via the communication section 809, and / or installed from a removable medium 811. The program code contained in the computer program can be transmitted using any suitable network medium, including but not limited to: wireless, wired, etc., or any suitable combination thereof.

[0133] In such an embodiment, the computer program can be downloaded and installed from a network via the communication section 809, and / or installed from the removable medium 811. When the computer program is executed by the processor 801, it performs the functions defined in the system of this application embodiment. According to the embodiments of this application, the systems, devices, apparatuses, modules, units, etc., described above can be implemented by computer program modules.

[0134] According to embodiments of this application, program code for executing the computer programs provided in the embodiments of this application can be written in any combination of one or more programming languages. Specifically, these computational programs can be implemented using high-level procedural and / or object-oriented programming languages, and / or assembly / machine languages. Programming languages ​​include, but are not limited to, languages ​​such as Java, C++, Python, "C", or similar programming languages. The program code can be executed entirely on the user's computing device, partially on the user's device, partially on a remote computing device, or entirely on a remote computing device or server. In cases involving remote computing devices, the remote computing device can be connected to the user's computing device via any type of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computing device (e.g., via the Internet using an Internet service provider).

[0135] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram or flowchart, and combinations of blocks in a block diagram or flowchart, may be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0136] Those skilled in the art will understand that the features described in the various embodiments of this application can be combined and / or combined in various ways, even if such combinations or combinations are not explicitly described in this application. In particular, the features described in the various embodiments of this application can be combined and / or combined in various ways without departing from the spirit and teachings of this application. All such combinations and / or combinations fall within the scope of this application.

Claims

1. A method for determining risk level, characterized in that, The method includes: Acquire target data, extract semantic features from the target data, and generate target feature semantic labels; The target feature semantic label is input into a pre-trained first risk model to generate a first risk label and a first risk value, wherein the first risk model is trained and generated based on first historical data, and the first historical data has a first risk level; The first risk label is input into a pre-trained second risk model to generate a second risk value, wherein the second risk model is trained based on second historical data, the second historical data has a second risk level, and the second risk level is lower than the first risk level; Based on the first risk value and the second risk value, generate a target risk value for the target data; and Based on the target risk value, the target risk level of the target data is determined. The target data includes target video data. The process involves acquiring the target data, extracting semantic features from the target data, and generating target feature semantic tags, including: Acquire target video data, perform frame extraction on the target video data, and generate target image data; The target image data is converted into text to generate target text data; and The target text data is subjected to semantic feature extraction using natural language processing techniques to generate the target feature semantic labels.

2. The method according to claim 1, characterized in that, The target image data is converted into text to generate target text data, including: Anomaly feature data is generated by extracting abnormal features from the target image data using a basic encoder; and The abnormal feature data is processed to generate the target text data.

3. The method according to claim 1, characterized in that, Based on the first risk value and the second risk value, a target risk value for the target data is generated, including: Preset the first and second weights; Calculate the product of the first risk value and the first weight to generate the first association value; Calculate the product of the second risk value and the second weight to generate the second correlation value; and The target risk value is generated based on the first correlation value and the second correlation value.

4. The method according to claim 3, characterized in that, Preset first and second weights, including: Calculate the difference between the first risk level and the second risk level; and Based on the difference between the first risk level and the second risk level, a first weight and a second weight are preset, wherein the difference between the first weight and the second weight is mapped to the difference between the first risk level and the second risk level.

5. The method according to claim 1, characterized in that, The method also includes: Based on the first risk level, the second risk level, and the difference between the first and second risk levels, calculate the shared loss value for the first risk model and the second risk model; and Based on the shared loss value, the first risk model and the second risk model are updated.

6. The method according to any one of claims 1 to 5, characterized in that, The method also includes: The first risk value is tested using the second risk value. If the test fails, a rollback is performed and a notification is sent.

7. A risk level determination device, characterized in that, The device includes: The first acquisition module is used to acquire target data, extract semantic features from the target data, and generate target feature semantic labels; The first generation module is used to input the target feature semantic label into a pre-trained first risk model to generate a first risk label and a first risk value, wherein the first risk model is trained and generated based on first historical data, and the first historical data has a first risk level. The second generation module is used to input the first risk label into a pre-trained second risk model to generate a second risk value, wherein the second risk model is trained and generated based on second historical data, the second historical data has a second risk level, and the second risk level is lower than the first risk level. The third generation module is used to generate a target risk value for the target data based on the first risk value and the second risk value; and The first determining module is used to determine the target risk level of the target data based on the target risk value. The target data includes target video data. The process involves acquiring the target data, extracting semantic features from the target data, and generating target feature semantic tags, including: Acquire target video data, perform frame extraction on the target video data, and generate target image data; The target image data is converted into text to generate target text data; and The target text data is subjected to semantic feature extraction using natural language processing techniques to generate the target feature semantic labels.

8. An electronic device, comprising: One or more processors; Memory, used to store one or more computer programs. The characteristic feature is that the one or more processors execute the one or more computer programs to implement the steps of the method according to any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program or instructions are executed by a processor, they implement the steps of the method according to any one of claims 1 to 6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.

Images