Marketing risk control system based on big data

Abstract

The invention discloses a marketing risk control system based on big data. The marketing risk control system comprises a server module, a client integration module and a device identification module. The server module and the client integration module realize a marketing risk control function through the following steps: constructing a server and a client; the server side generates data, the client side generates a fingerprint after receiving the data, and the server side generates an equipment uniqueness identifier according to the fingerprint and the data; obtaining operation time of a user, generating a time sequence by the operation time, extracting features of the time sequence, and constructing a behavior risk probability model according to the features of the time sequence; calculating a risk value according to the equipment uniqueness identifier; random disturbance data is added in a server, and high-precision user identification and cheating detection are realized by detecting information of clients, identifying cheating clients and rendering unique identifiers and behavior time sequence modeling through hardware, so that fairness and safety of marketing activities are protected.

Description

Technical Field

[0001] The present invention relates to the technical field of computer software, and in particular, to a marketing risk control system based on big data. Background Art

[0002] With the rapid development of Internet technology, marketing activities have become increasingly digital and networked, and various online marketing means have emerged in an endless stream, which has damaged the interests of merchants and consumers. In order to effectively combat these cheating behaviors and ensure the fairness and security of marketing activities, marketing risk control technology has emerged as the times require.

[0003] For example, Chinese Patent Application No. 202210079401.5 discloses a marketing anti-cheating system based on big data, which is applied to marketing anti-cheating. The system includes a cloud server and a user terminal; when the system works, the marketing anti-cheating function is realized through the following steps: when a user performs an operation related to a marketing activity on the user terminal, a risk identification request is sent to the cloud server, and after receiving the request, the cloud server calls the data collection module to send a data collection instruction to the user terminal; the user terminal responds to the data collection instruction, and after receiving the data, the cloud server calls the risk identification control module to perform risk judgment by using the methods of comparing with a black and white list database and judging by a risk model respectively; the cloud server obtains a risk judgment instruction; the cloud server transmits the risk judgment instruction as a response to the risk identification request to the user terminal, and the user terminal accepts or rejects the user operation according to the risk judgment instruction.

[0004] In the existing patent technology of marketing risk control systems, device identification mainly relies on the hardware information of devices, such as MAC addresses, IMEI numbers, etc. However, these hardware information are easily tampered with or forged, resulting in a serious impact on the accuracy of device identification. In addition, with the development of virtualization technology and simulators, cheaters can use these technologies to simulate a real device environment, further bypassing the detection of device identification. Therefore, how to improve the accuracy of device identification and the anti-bypass ability has become an important challenge faced by marketing risk control technology. Summary of the Invention

[0005] The present application provides a marketing risk control system based on big data, which realizes high-precision user identification and cheating detection through unique identification of hardware rendering and behavior time series modeling, and protects the fairness and security of marketing activities.

[0006] The present application provides a marketing risk control system based on big data. The marketing risk control system includes a server module, a client integration module, and a device identification module; the server module and the client integration module realize the function of marketing risk control through the following steps: S101, construct a server and a client; S102, The server generates data. After the client receives the data, it generates a fingerprint. Based on the fingerprint and the data, the server generates a device uniqueness identifier; S103, Obtain the user's operation time. The operation time generates a time series, and extract the features of the time series. Based on the features of the time series, construct a behavioral risk probability model; S104, Calculate the risk value based on the device uniqueness identifier; S105, Add random perturbation data in the server. By detecting the information of the client, identify cheating clients.

[0007] Preferably, the time series features include time-series behavior features and interaction biometric features. For the time-series behavior features, through the client integration module, monitor the user's operation events and their timestamps, and generate a time series in sequence according to the obtained user operation time; for the interaction biometric features, the interaction behavior features are extracted by the average acceleration of mouse movement and the entropy of touch pressure.

[0008] Preferably, the average acceleration of mouse movement is used to measure the dynamic characteristics of the user's mouse movement. Calculate the average acceleration of mouse movement through the velocity vector. The formula is: , where, represents the average value of acceleration, represents the number of acceleration values. Since the acceleration is calculated starting from the second velocity vector, there are M - 1 acceleration values in total, represents the index of the acceleration vector, represents the acceleration.

[0009] Preferably, the entropy of touch pressure is used to measure the randomness of the user's touch pressure distribution. Calculate the entropy of touch pressure according to the probability distribution of the interval. The formula is: , where, represents the entropy of touch pressure, K represents the number of intervals into which the pressure value is divided, represents the index of the pressure value interval, represents the probability that the pressure value falls within the kth interval, represents the logarithm of.

[0010] Preferably, calculate the risk value based on the device uniqueness identifier. The formula is: , where, is the comprehensive risk value, is the number of times the current device has received, is the maximum number of times allowed by the platform, and ≠0, is the probability of user cheating, is the IP address risk score, and the value range is between [0, 1], , is a weight coefficient and satisfies + = 1.

[0011] Preferably, the method for compensating for differences in different browsers in the client integration module is as follows: S201. Set the standardized generated content according to the software of the client. S202. Place the set standardized generated content at the center of the visible area. S203. Use the software for off-screen rendering. S204. Compensate for the differences according to the differences in rendering colors among different browsers.

[0012] Preferably, calculate the coordinates of the standardized generated content in the visible area based on the width and height of the visible area and the width and height of the rendered content. The formula is: , , where and represent the coordinates of the standardized generated content in the visible area, represents the width of the visible area, represents the height of the visible area, represents the width of the rendered content, represents the height of the rendered content.

[0013] Preferably, the method for the device recognition module to recognize the device is as follows: S301. Measure the first frame rendering time, divide the performance of the device according to the first frame rendering time, and perform dynamic grading according to the device performance. S302. Extract multi-dimensional features. S303. Generate a dynamic ID according to the device load condition, detect the visualization area, and identify the virtual environment through the physical GPU characteristics. S304. Recognize the device through multi-dimensional features, dynamic ID, detection of the visualization area, and identification of the virtual environment.

[0014] Preferably, divide the performance of the device according to the first frame rendering time, set a first threshold and a second threshold, regard the device with the first frame rendering time less than the first threshold as a high-performance device, regard the device with the first frame rendering time greater than or equal to the first threshold and less than the second threshold as a balanced device, and regard the device with the first frame rendering time greater than or equal to the second threshold as a low-performance device.

[0015] Preferably, according to the current load condition of the device, the devices are divided into different load levels. Different device IDs are generated according to the load levels of the devices. By embedding the load level information in the device unique identifier, different IDs are generated for the same device under different load levels, but the IDs generated multiple times under the same load level remain consistent.

[0016] One or more technical solutions provided in this application have at least the following technical effects or advantages: Through hardware rendering of unique identifiers and behavioral timing modeling, high-precision user identification and cheating detection are achieved, protecting the fairness and security of marketing activities. The device-level uniqueness error rate decreases, and the detection rate of behavioral anomalies increases, effectively intercepting cheating behaviors while reducing the misjudgment of normal users. By combining device rendering of unique identifiers and behavioral timing modeling, higher user identification accuracy and cheating detection rate are provided. Anti-bypass enhancement designs such as dynamic noise injection and virtualization environment detection are introduced, which can continuously counter new attack methods. By reducing the misjudgment of normal users, the solution improves the user experience and the reputation of the platform. Eliminate browser difference interference, ensure that the same device generates consistent features in different browsers, maintain high uniqueness, and accurately identify hardware differences in a cross-browser environment to avoid misjudgment. By standardizing content generation parameters, off-screen rendering and center positioning, browser difference compensation and other measures, the cross-browser consistency is significantly improved, and the misjudgment rate caused by browser differences is reduced. Off-screen rendering and center positioning achieve isolation of the rendering environment, are not affected by page layout, and improve the anti-interference ability of the system. Through dynamic load grading, multi-reference collaboration and privacy-security balance design, stable hardware identification across devices and resolutions is achieved, while enhancing privacy protection and anti-attack capabilities. Through physical GPU verification, the discrimination and security of hardware identification are significantly improved. Dynamic rendering complexity grading ensures a good user experience on devices with different performances. Anti-virtualization reinforcement measures effectively intercept virtualization attacks, improve the security of hardware identification, achieve cross-device hardware identification with high discrimination and low performance consumption, enhance the identification ability of virtualization environments, and improve the security and accuracy of hardware identification. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 It is a block diagram of a marketing risk control system based on big data according to the present invention; Figure 2 It is a schematic flow chart of the marketing risk control function implemented by the server module and the client integration module of the present invention; Figure 3 It is a schematic flow chart of compensating for different browser differences in the client integration module of the present invention; Figure 4 It is a schematic flow chart of the device identification module of the present invention for identifying devices. Detailed implementation mode

[0018] To facilitate the understanding of the present invention, the present application will be described more comprehensively below with reference to the relevant drawings; the preferred embodiments of the present invention are shown in the drawings. However, the present invention can be implemented in many different forms and is not limited to the embodiments described herein; on the contrary, these embodiments are provided to make the disclosure of the present invention more thoroughly understood.

[0019] It should be noted that the terms "vertical", "horizontal", "up", "down", "left", "right" and similar expressions used herein are for illustrative purposes only and do not represent the only embodiment.

[0020] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the technical field to which the present invention belongs; the terms used in the description of the present invention herein are only for the purpose of describing specific embodiments and are not intended to limit the present invention; the term "and / or" used herein includes any and all combinations of one or more of the related listed items.

[0021] Embodiment 1: Figure 2 It is a process schematic diagram of a marketing risk control system based on big data according to an embodiment of the present invention, applied to marketing risk control, as Figure 1 shown, the system includes a server module, a client integration module and a device identification module. The server module is connected to the client integration module, and both the server module and the client integration module are connected to the device identification module. When the system works, the function of marketing risk control is realized through the following steps: S101, construct the server and client of marketing risk control; Specifically, for the server module, the server of marketing risk control uses the cloud service Alibaba Cloud and installs a database system with MySQL. The MySQL is suitable for structured data storage, such as device unique ID, user information, transaction records, etc. MySQL has high performance, reliability and ease of use, and is suitable for processing a large amount of transactional data. Configure and optimize the database according to business requirements, including index design, partitioning strategy, read-write separation, etc., to improve data access speed and system stability; for the client integration module, integrate the software development kit (SDK) into the client. The SDK includes functional modules such as WebGL rendering, feature extraction, hash generation, behavior monitoring and risk assessment.

[0022] S102, the server generates data, the client generates a fingerprint after receiving the data, and the server generates a device uniqueness identifier according to the fingerprint and the data; Further, the server generates rendering data based on the task space, and the rendering data includes shape category, color gradient, rotation angle, and graphic complexity. For the shape category, the shape category is the geometric shape of the specified rendering graphic, which is randomly selected from a predefined set of shapes. For the color gradient, the color gradient defines the color change range of the rendering graphic, and the RGB values of the starting color and the ending color are randomly generated, with the value of each channel ranging from 0 to 255. For the rotation angle, the rotation angle is the rotation angle of the specified graphic around the center point, and an angle value is randomly generated from a specified range. For the graphic complexity, the graphic complexity controls the complexity of the graphic. For polygons, the number of vertices is randomly generated, and for curves, the number of control points or the order is randomly generated. The client receives the generated rendering data through the communication interface with the server and uses the received rendering data to start the rendering engine to generate the corresponding image.

[0023] The client integration module generates an anti-aliasing fingerprint and a floating-point precision fingerprint based on the rendered image. For the anti-aliasing fingerprint, the pixel data of the image after rendering is obtained, and the pixel data is defined as a matrix. The Sobel operator, an edge detection algorithm, is used to identify the edge pixels in the image. Another pixel is selected within the neighborhood of the slightly offset edge pixel for comparison. The slight offset is a very small displacement, usually one or a few pixel units. The color value difference between the edge pixel and the pixel after offset in the same color channel is calculated, and the color value difference is obtained through absolute value calculation. The color differences of all edge pixels and all color channels are summed to obtain a total sum. Then, the total sum is divided by the total number of edge pixels to obtain the average color difference, and this average value is the anti-aliasing edge feature vector. The obtained feature vector is hashed to generate the anti-aliasing fingerprint. For the WebGL floating-point precision fingerprint, in the WebGL shader, the vertex positions are slightly perturbed to reflect the differences in the floating-point operation precision of the hardware. The perturbed vertex positions are obtained based on the original vertex positions. Based on the deviation between the perturbed vertex positions and the original vertex positions, a deviation matrix is formed. The deviation matrix is transposed, and the transposed deviation matrix is multiplied by the deviation matrix. The multiplied matrix is subjected to eigenvalue decomposition, and the maximum eigenvalue and the minimum eigenvalue are extracted from the results of the eigenvalue decomposition. The eigenvalues are hashed to generate the floating-point precision fingerprint.

[0024] Record the time required for the client to complete the rendering task, concatenate the anti-aliasing hash value, floating-point precision hash value, and rendering time consumption, perform a hash process on the concatenated data string to generate a hash value of a fixed length, which is the unique device identifier. The concatenation is to connect these feature values in a certain order into a long data string. The hash process is an algorithm that converts input data of any length into an output of a fixed length (hash value), and it is almost impossible for different input data to produce the same hash value.

[0025] S103, obtain the operation time of the user, generate a time series from the operation time, extract the features of the time series, and construct a behavioral risk probability model based on the features of the time series; Specifically, the time series features include temporal behavior features and interaction biometric features. For the temporal behavior features, by using the client application, monitor and record the user's operation events (such as clicks, swipes, inputs, etc.) and their timestamps, generate a time series in order according to the obtained user operation time, calculate the time difference between adjacent operation events according to the time series, and calculate the variance of the operation time interval according to the time difference. For the interaction biometric features, the interaction behavior features are extracted through the mean value of mouse movement acceleration and touch pressure entropy. The mean value of mouse movement acceleration is used to measure the dynamic characteristics of the user's mouse movement, reflecting the speed change and amplitude of the user's mouse operation. In the client application, monitor and record the speed vector of the mouse movement. The speed vector is calculated by recording the change amount of the mouse position and the time interval, store the speed vector in a sequence in time order, and calculate the mean value of mouse movement acceleration through the speed vector. The formula is: , where represents the average value of acceleration, represents the number of acceleration values. Since the acceleration is calculated starting from the second speed vector, there are M - 1 acceleration values in total, represents the index of the acceleration vector, represents acceleration. The touch pressure entropy is used to measure the randomness of the user's touch pressure distribution, reflecting the diversity and uncertainty of the user's touch pressure. According to the range and distribution of the pressure values, divide the pressure value intervals, calculate the probability distribution of the intervals, and calculate the touch pressure entropy according to the probability distribution of the intervals. The formula is: , where represents the touch pressure entropy, used to measure the randomness of the user's touch pressure distribution. K represents the number of intervals into which the pressure values are divided, that is, the entire range of pressure values is divided into K non-overlapping sub-intervals, represents the index of the pressure value interval, represents the probability that the pressure value falls within the kth interval, that is, the proportion of the number of pressure values in the kth interval to the total number of pressure values, Represents the logarithm of

[0026] Construct a behavioral risk probability model based on the variance of the operation time interval, the mean value of the mouse movement acceleration, and the touch pressure entropy. The formula in the risk probability model is: where is the probability of user cheating, is the variance of the operation time interval, represents the average value of the acceleration, represents the touch pressure entropy, , and and are model coefficients obtained from training data. Train and test the behavioral risk probability model. According to the behavioral data of new users, extract the time-series behavioral features, and input the extracted features into the trained behavioral risk probability model. The model outputs the probability of user cheating.

[0027] S104, calculate the risk value according to the device identifier, the probability of user cheating, and the user behavior; Furthermore, collect the number of times the current device has received rewards from the server database or real-time logs, and calculate the risk value based on the device identifier, the probability of user cheating, and the user behavior. The formula is: where is the comprehensive risk value, used to measure the risk level of the user, is the number of times the current device has received rewards, indicating the number of times the device has received rewards or discounts on the platform, is the maximum number of times the platform allows to receive rewards, and ≠0, is a threshold set by the platform to limit the number of times the device receives rewards or discounts, is the probability of user cheating, is the IP address risk score, with a value range between [0, 1], calculated based on historical cheating records, used to measure the risk level of the IP address. 0 indicates no risk, and 1 indicates high risk. , are weight coefficients, used to adjust the contribution degree of each factor in the comprehensive risk value, and satisfy + = 1. Set dynamic hierarchical interception according to the calculated risk value. Set the hierarchical interception threshold based on business requirements and risk control strategies. When R < 0.3, it is considered that the user risk is low, and the user request is directly allowed; when 0.3 ≤ R < 0.7, it is considered that the user risk is medium, and additional verification steps (such as SMS verification, face recognition, etc.) are required before deciding whether to allow the user request; when R ≥ 0.7, it is considered that the user risk is high, and the user request is directly intercepted.

[0028] S105. Add random perturbation data to the data generated on the server side, and identify cheating clients running in a virtualized environment by detecting client information. Specifically, introduce random perturbation into the rotation angle data, generate uniformly distributed random numbers, and add them to the original angle to obtain new rotation angle data. By introducing random noise, the result of the rendering task becomes uncertain, increasing the difficulty for cheaters to predict or simulate the rendering result. Due to the randomness of the noise, it is difficult for cheaters to bypass the system detection through fixed patterns or algorithms, thus improving the anti-bypass ability of the system. When the client is running, obtain the GPU driver information. By detecting the GPU driver information, clients running in a virtualized environment can be accurately identified. For the identified clients running in a virtualized environment, the system will reject their requests or execute other restrictive measures, thus effectively preventing cheating behavior from occurring.

[0029] The technical solutions in the embodiments of the present application at least have the following technical effects or advantages: Through the unique identifier of hardware rendering and behavior timing modeling, high-precision user identification and cheating detection are achieved, protecting the fairness and security of marketing activities. The device-level unique error rate decreases, and the detection rate of abnormal behaviors increases. Cheating behaviors are effectively intercepted, while reducing the misjudgment of normal users. By combining the unique identifier of device rendering and behavior timing modeling, higher user identification accuracy and cheating detection rate are provided. Anti-bypass enhancement designs such as dynamic noise injection and virtualized environment detection are introduced to continuously counter new attack methods. By reducing the misjudgment of normal users, the user experience and the reputation of the platform are improved.

[0030] Embodiment 2: On the basis of Embodiment 1, in this embodiment, by selecting positions and compensating for differences in different browsers, the generated content has the same rendering result in different browsers, reducing the decrease in the stability of device fingerprints, as Figure 3 shown.

[0031] S201. Set standardized generated content according to the software of the client. Furthermore, the standardized generated content includes geometric shape standardization, dynamic effect standardization, and rendering environment isolation. For geometric shape standardization, basic shapes (circles, squares) are used, and complex curves (such as Bezier curves) are avoided to reduce shape inconsistencies caused by differences in browser rendering engines. The size is defined in pixels, and the size value is an even number to reduce edge blurring or differences caused by anti-aliasing processing, ensuring that the graphics have the same size and appearance on different browsers and devices. The sRGB standard color values are used, and the color management function of the browser is disabled during rendering to ensure that the color values are directly applied without any conversion, making the colors appear consistent on different browsers and devices and avoiding color deviations caused by color management. For dynamic effect standardization, the animation frame rate is set to 60 FPS to ensure the same smoothness of the animation on different devices. The rotation angle increment per frame is 1°, maintaining the consistency of the rotation speed. By setting CSS properties, unnecessary optimizations of the browser for the animation are avoided, which may affect the consistency of the animation, ensuring that the rotation animation has the same rotation speed and smoothness on different browsers and devices. The linear gradient direction is fixed to horizontal or vertical, and diagonal gradients are avoided because diagonal gradients cause rendering inconsistencies due to coordinate system differences, ensuring that the gradient effect appears consistent on different browsers and devices. For rendering environment isolation, forced use of high-precision floating-point operations is set in the WebGL configuration to reduce rendering problems caused by differences in floating-point precision, ensuring that the graphics have the same rendering precision and detail performance on different browsers and devices. The rendering buffer is set to be retained in the WebGL configuration so that the rendering result can be reread or modified when needed, providing additional control over the rendering result and ensuring consistent access and modification of the rendering buffer on different browsers and devices.

[0032] S202, Place the set standardized generated content at the center of the visible area; Specifically, use JavaScript to obtain the width and height of the visible area of the current browser window. The width and height of the visible area represent the area inside the browser window available for displaying content. The width and height of the content to be rendered are preset, and the coordinates of the standardized generated content in the visible area are calculated based on the width and height of the visible area and the width and height of the rendered content. The formula is: , , where, and represent the coordinates of the standardized generated content in the visible area, represents the width of the visible area, represents the height of the visible area, represents the width of the rendered content, Represents the height of the rendered content. When using Canvas for drawing, apply the calculated x and y coordinates to the content to be rendered.

[0033] S203, perform off-screen rendering using software; Furthermore, use the software JavaScript constructor to create an off-screen Canvas object with a width of 100 pixels and a height of 100 pixels. Use the WebGL rendering method to obtain the rendering context of the off-screen Canvas. This context will be used for actual drawing operations in the background thread, including drawing graphics, applying styles, etc. Create a script file and define the rendering logic in it. The script file will not block the execution of the main thread. In the script file, receive the reference to the off-screen Canvas passed from the main thread and perform drawing operations using the obtained rendering context. These operations are completely carried out in the background thread and are not affected by factors such as CSS styles and page scaling in the main thread. After completion of the drawing, the script file sends the image data back to the main thread and converts the image data into an image to be displayed on the page.

[0034] S204, compensate for the differences in rendering colors among different browsers; Specifically, draw a standard checkerboard pattern on the page. The pattern consists of black and white squares and serves as a reference for benchmark testing to compare the rendering results of different browsers. Use Canvas to obtain the pixel data of a specific area (such as a white square or a black square) in the checkerboard pattern, extract the color values of the red (R), green (G), and blue (B) channels, which are the actual measured color values. Set the standard color values, where the standard color values represent the colors of the squares in the checkerboard pattern under ideal conditions. Calculate the differences between the actual measured color values and the standard color values, and calculate the color compensation matrix based on the color differences. The form of the matrix is as follows: , where represents the color compensation matrix, ΔR represents the color difference in the red channel, ΔG represents the color difference in the green channel, and ΔB represents the color difference in the blue channel. respectively represent the standard color values of the red, green, and blue channels. Obtain information such as the type and version of the browser, set up a compensation library that contains compensation parameters for different browser types and versions. According to the browser type and version number, query the corresponding compensation parameters in the compensation library, and dynamically adjust the rendering data based on the queried compensation parameters and the color compensation matrix.

[0035] The technical solutions in the embodiments of the present application at least have the following technical effects or advantages: eliminating browser difference interference, ensuring consistent features are generated on the same device in different browsers, maintaining high uniqueness, accurately identifying hardware differences in a cross-browser environment, avoiding misjudgment, and significantly improving cross-browser consistency and reducing the misjudgment rate caused by browser differences through measures such as standardizing content generation parameters, off-screen rendering and center positioning, and browser difference compensation. Off-screen rendering and center positioning achieve isolation of the rendering environment, are not affected by page layout, and improve the anti-interference ability of the system.

[0036] Embodiment 3: Based on Embodiment 2 above, fixed rendering data may cause lags and blurred effects on devices with different performances. In this embodiment, by dynamically grading the rendering complexity, a good user experience is ensured on devices with different performances, as Figure 4 shown.

[0037] S301, measure the first-frame rendering time, divide the performance of the device according to the first-frame rendering time, and perform dynamic grading according to the device performance; Furthermore, measure the first-frame rendering time, divide the device types according to the first-frame rendering time, set a first threshold and a second threshold. Devices with a first-frame rendering time less than the first threshold are regarded as high-performance devices, devices with a first-frame rendering time greater than or equal to the first threshold and less than the second threshold are regarded as balanced devices, and devices with a first-frame rendering time greater than or equal to the second threshold are regarded as low-performance devices. Perform dynamic grading according to the performance type of the device. For high-performance devices, set the highest fractal iteration times and the number of particles, and set the anti-aliasing level to 8 times. For balanced devices, reduce the fractal iteration times and the number of particles on the basis of high-performance devices, and set the anti-aliasing level to 4 times. For low-performance devices, further reduce the fractal iteration times and the number of particles on the basis of balanced devices, and do not set the anti-aliasing level. Re-measure the rendering time every 5 seconds. If the rendering time exceeds the current mode threshold for 3 consecutive times, switch to the next-level mode.

[0038] S302, extract multi-dimensional features; The multi-dimensional features include spatial features, temporal features, and browser rendering strategies. For spatial features, the original gradient values of the edge regions are extracted from the rendered graphics. The gradient values reflect the degree of change in pixel values in the image. The original gradient values are normalized according to the different performances of the devices, and the normalized gradient values are used to perform anti-aliasing processing on the graphics edges. Then, the edge regions are measured, and the attenuation characteristics of the edge regions are analyzed according to the measurement results. The ESI value is calculated based on the attenuation data. The ESI value is inversely proportional to the edge attenuation degree, that is, the sharper the edge (the slower the attenuation), the higher the ESI value. The calculated ESI value is stored; for temporal features, the current frame rate during the rendering process is recorded, the variance is calculated based on the frame rate, and the calculated variance is stored. In high-performance devices, the timer function of WebGL is used to record the start and end times of the GPU and CPU. According to the recorded times, the cooperation delay between the GPU and CPU is calculated. Delay difference = GPU completion time - CPU start time, and the delay difference is stored. The browser rendering strategy has been described in the second embodiment above and will not be elaborated in this embodiment.

[0039] S303. Generate a dynamic ID according to the device load condition, detect the visualization area, and identify the virtual environment through the physical GPU characteristics; Specifically, according to the current load condition of the device (such as CPU usage rate, memory occupancy rate, etc.), the devices are divided into different load levels (such as low load, medium load, high load). Different device IDs are generated according to the load levels of the devices, which is achieved by embedding load level information in the device identifier or using different hash algorithms, ensuring that the device IDs generated under different load levels are unique and stable, that is, the same device should generate different IDs under different load levels, but the IDs generated multiple times under the same load level are consistent. The dynamic ID generation mechanism can ensure that unique and stable device identifiers can be generated in different modes; when the device is initialized or the user accesses for the first time, the hash value of the visualization area is calculated (such as based on parameters such as the size and resolution of the visualization area). When each request is made, the hash value of the visualization area is recalculated, and the recalculated hash value is compared with the initial hash value, and the deviation is calculated. If the deviation exceeds the threshold, a risk flag is triggered, indicating that the size of the visualization area has been tampered with; the same sphere collision detection is run multiple times on the physical GPU, and the mean and standard deviation of the collision point coordinates are statistically calculated as the expected results. The same ray tracing simulation calculation is run on the device to be verified to obtain the simulation results, and the difference between the simulation results and the expected results is calculated. If the difference is greater than the preset threshold, it is marked as a virtual environment, otherwise, it is not a virtual environment.

[0040] S304. Identify the device through multi-dimensional features, dynamic ID, detection of the visualization area, and identification of the virtual environment.

[0041] The technical solutions in the embodiments of the present application at least have the following technical effects or advantages: Through dynamic load grading, multi-reference collaboration, and privacy-security balance design, stable hardware recognition across devices and resolutions is achieved. At the same time, privacy protection and anti-attack capabilities are enhanced. Through physical GPU verification, the discrimination and security of hardware recognition are significantly improved. The dynamic rendering complexity grading ensures a good user experience on devices with different performance. The anti-virtualization reinforcement measures effectively intercept virtualization attacks, improving the security of hardware recognition, achieving cross-device hardware recognition with high discrimination and low performance consumption, enhancing the recognition ability for virtualized environments, and improving the security and accuracy of hardware recognition.

[0042] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention may have various changes and modifications. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A marketing risk control system based on big data, characterized in that, The marketing risk control system includes a server module, a client integration module, and a device identification module; the server module and the client integration module implement the functions of marketing risk control through the following steps: S101, construct the server and the client; S102, the server generates data, and after the client receives the data, it generates a fingerprint. Based on the fingerprint and the data, the server generates a device uniqueness identifier; S103, obtain the operation time of the user, generate a time series from the operation time, extract the features of the time series, and construct a behavioral risk probability model based on the features of the time series; S104, calculate the risk value according to the device uniqueness identifier; S105, add random perturbation data in the server, and identify cheating clients by detecting the information of the client.

2. The marketing risk control system based on big data according to claim 1, wherein The time series features include temporal behavior features and interaction biometric features. For the temporal behavior features, through the client integration module, monitor the user's operation events and their timestamps, and generate a time series in order according to the obtained user operation time; for the interaction biometric features, the interaction behavior features are extracted by the average acceleration of mouse movement and the touch pressure entropy.

3. The marketing risk control system based on big data according to claim 2, characterized in that, The average value of the mouse movement acceleration is used to measure the dynamic characteristics of the user's mouse movement. The average value of the mouse movement acceleration is calculated through the velocity vector, and the formula is: , where represents the average value of the acceleration, represents the number of acceleration values. Since the acceleration is calculated starting from the second velocity vector, there are a total of M - 1 acceleration values, represents the index of the acceleration vector, represents the acceleration.

4. The marketing risk control system based on big data according to claim 2, wherein The touch pressure entropy is used to measure the randomness of the user's touch pressure distribution. The touch pressure entropy is calculated according to the probability distribution of intervals, and the formula is: , where represents the touch pressure entropy, K represents the number of intervals into which the pressure values are divided, represents the index of the pressure value interval, represents the probability that the pressure value falls within the k-th interval, represents the logarithm of.

5. A marketing risk control system based on big data according to claim 1, characterized in that, Calculate the risk value based on the unique device identifier. The formula is: , where is the comprehensive risk value, is the current number of times the device has been received, is the maximum number of times the platform allows to receive, and ≠ 0, is the probability of user cheating, is the IP address risk score, and the value range is between [0, 1], , are the weight coefficients, and satisfy + = 1.

6. The marketing risk control system based on big data according to claim 1, characterized in that The method for compensating for differences in different browsers in the client integration module is as follows: S201, set the standardized generated content according to the software of the client; S202, place the set standardized generated content at the center of the visible area; S203, perform off-screen rendering using software; S204, compensate for the differences according to the differences in rendering colors of different browsers.

7. The marketing risk control system based on big data according to claim 6, wherein, Calculate the coordinates of the normalized generated content in the visible area based on the width and height of the visible area and the width and height of the rendered content. The formula is: , , where and represent the coordinates of the normalized generated content in the visible area, represents the width of the visible area, represents the height of the visible area, represents the width of the rendered content, represents the height of the rendered content.

8. The marketing risk control system based on big data according to claim 1, characterized in that, The method for the device identification module to identify the device is as follows: S301, measure the first frame rendering time, divide the performance of the device according to the first frame rendering time, and perform dynamic grading according to the device performance; S302, extract multi-dimensional features; S303, generate a dynamic ID according to the device load condition, detect the visualization area, and identify the virtual environment through the physical GPU characteristics; S304, identify the device through multi-dimensional features, dynamic ID, detection of the visualization area, and identification of the virtual environment.

9. The marketing risk control system based on big data according to claim 8, characterized in that, Divide the performance of the device according to the first frame rendering time, set a first threshold and a second threshold, regard the device with the first frame rendering time less than the first threshold as a high-performance device, regard the device with the first frame rendering time greater than or equal to the first threshold and less than the second threshold as a balanced device, and regard the device with the first frame rendering time greater than or equal to the second threshold as a low-performance device.

10. A marketing risk control system based on big data as claimed in claim 8, wherein, According to the current load condition of the device, divide the device into different load levels. According to the load level of the device, generate different device IDs. By embedding the load level information in the device unique identifier, the same device generates different IDs under different load levels, but the IDs generated multiple times under the same load level remain the same.

Images