AI-generated content front-end real-time identification method and system based on dual-mode verification

Abstract

The embodiment of the specification discloses an AI-generated content front-end real-time identification method and system based on dual-mode verification. Among them, the AI-generated content front-end real-time identification method based on dual-mode verification includes obtaining user file information and user operation data based on the front-end environment running on the user equipment, and generating behavior fingerprints according to the user operation data; the edge density processing and the word frequency jump detection processing are performed on the user file information to obtain the content features of the user file information; based on the preset dual-mode risk judgment engine, the risk score data of the user is determined according to the content features and the user operation data; when the risk score data satisfies the first preset condition, the user file information is determined as non-AI generated content, and the behavior fingerprints are added to the metadata corresponding to the user file information to obtain the behavior fingerprint metadata. The embodiment of the specification improves the content detection accuracy while reducing the consumption of computing resources.

Description

Technical Field

[0001] One or more embodiments of this specification relate to the field of data security technology, specifically to a method and system for real-time front-end recognition of AI-generated content based on bimodal verification. Background Technology

[0002] With the rapid development of artificial intelligence technology, AI-generated content (such as deepfake images and machine-generated text) is spreading increasingly widely on the internet. The constantly evolving technologies for generating this content make it increasingly difficult to distinguish from real content, posing a significant challenge to verifying the authenticity of information. For example, on social media platforms, fake AI-generated images or text may mislead users or even be used for malicious purposes, such as spreading misinformation or online fraud. Currently, detection technologies for AI-generated content mainly fall into two categories: content feature analysis technology and user authentication technology. Content feature analysis technology analyzes the statistical characteristics of a document by training AI models (such as convolutional neural networks and natural language processing models). However, this technology relies solely on document features and cannot distinguish between "rapidly generated content created by a real person" and "machine-generated content," making it easily bypassed by carefully designed AI-generated content. Furthermore, content analysis typically requires the deployment of large AI models, resulting in high computational costs and making it difficult to run in front-end environments such as browsers. In addition, user authentication technology indirectly judges user credibility and infers the authenticity of content through account reputation ratings, mobile phone verification, and biometric authentication. However, this method also has obvious drawbacks: authentication technology is easily bypassed by identity forgery; for example, unauthorized individuals can purchase real-name accounts to publish false content. Furthermore, existing detection technologies also pose privacy risks. Cloud-based detection requires uploading user files, which carries the risk of data leakage, such as uploading ID card images to third-party servers. Secondly, they lack real-time performance; cloud-based detection has delays, affecting user experience. For example, image uploads may experience lag on social media platforms.

[0003] Existing technologies for detecting AI-generated content suffer from drawbacks such as limited detection dimensions, high computational costs, significant privacy risks, and poor real-time performance, making it difficult to meet the demands of practical applications for high accuracy, low resource consumption, strong privacy protection, and rapid response. Therefore, there is an urgent need for a real-time front-end recognition method for AI-generated content that can improve recognition accuracy. Summary of the Invention

[0004] This specification provides an embodiment of a method and system for real-time front-end recognition of AI-generated content based on bimodal verification. The technical solution is as follows: In a first aspect, embodiments of this specification provide a real-time front-end recognition method for AI-generated content based on bimodal verification, comprising: acquiring user file information and user operation data based on the front-end environment running on the user device, and generating a behavioral fingerprint based on the user operation data; performing edge density processing and word frequency jump detection processing on the user file information to obtain the content features of the user file information; determining the user's risk score data based on the content features and user operation data using a preset bimodal risk judgment engine; and when the risk score data meets a first preset condition, determining the user file information as non-AI-generated content, and adding the behavioral fingerprint to the metadata corresponding to the user file information to obtain behavioral fingerprint metadata.

[0005] Secondly, embodiments of this specification provide a front-end real-time recognition system for AI-generated content based on bimodal verification, comprising: a data acquisition module, used to acquire user file information and user operation data based on the front-end environment running on the user's device, and generate a behavioral fingerprint based on the user operation data; a feature extraction module, used to perform edge density processing and word frequency jump detection processing on the user file information to obtain the content features of the user file information; a risk determination module, used to determine the user's risk score data based on the content features and user operation data according to a preset bimodal risk determination engine; and a metadata tagging module, used to determine the user file information as non-AI-generated content when the risk score data meets a first preset condition, and add the behavioral fingerprint to the metadata corresponding to the user file information to obtain behavioral fingerprint metadata.

[0006] The beneficial effects of the technical solutions provided in some embodiments of this specification include at least the following: The embodiments in this specification employ a dual-modal verification method that combines user behavior and content. By comprehensively considering user operating habits and file content characteristics, the detection accuracy is improved. Machine-generated content is difficult to simultaneously forge human operating habits and natural content characteristics. Therefore, the embodiments in this specification can more effectively identify AI-generated content and reduce the false positive rate through dual-modal verification.

[0007] Moreover, the embodiments in this specification run on the front end, without relying on large AI models. Lightweight detection is achieved through a rule engine, which greatly reduces the consumption of computing resources and avoids network latency and data transmission costs caused by cloud computing.

[0008] Furthermore, the detection latency in this embodiment is reduced from 1200ms in cloud detection to 80ms, and the response time is less than 100ms, enabling rapid real-time interception. Eliminating network transmission and cloud queuing time eliminates noticeable lag when users upload images or text, greatly improving the user experience. This is particularly suitable for applications with high real-time requirements, such as content publishing on social media platforms.

[0009] Furthermore, the embodiments described in this specification can run locally, with user actions and file content processed locally without uploading to the cloud, effectively avoiding the risk of data leakage. Relying on native browser APIs, it complies with privacy regulations such as GDPR, fully protecting user privacy and data security.

[0010] Furthermore, the embodiments in this specification, through comprehensive analysis of user behavior and content characteristics, require attackers to simultaneously forge content features and simulate human operational behavior, significantly increasing the cost of forgery. Uploaders of AI-generated content typically exhibit characteristics of "high content quality + low operational complexity," while real users often employ multiple adjustments and random operational paths. The design of this invention can effectively identify this difference, thereby enhancing the ability to recognize AI-generated content and reducing the risk of being bypassed.

[0011] The embodiments in this specification solve the problems of single detection dimension, high computational cost, high privacy risk and poor real-time performance in the existing technology through innovative dual-modal verification technology. It provides an efficient, accurate, safe and real-time solution for the recognition of AI-generated content, which has broad application prospects and important practical significance. Attached Figure Description

[0012] To more clearly illustrate the technical solutions in the embodiments of this specification, the accompanying drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this specification. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0013] Figure 1 This is a schematic diagram illustrating an application scenario of a real-time front-end recognition method for AI-generated content based on bimodal verification, as provided in this manual.

[0014] Figure 2 This is a flowchart illustrating a real-time front-end recognition method for AI-generated content based on bimodal verification, as provided in this specification.

[0015] Figure 3 This is a flowchart illustrating the process of generating behavioral fingerprints based on user operation data, as provided in this manual.

[0016] Figure 4 This is a flowchart illustrating the process of obtaining content features of user file information as provided in this manual.

[0017] Figure 5 This is a schematic diagram of the structure of a front-end real-time recognition system for AI-generated content based on bimodal verification, as provided in this specification. Detailed Implementation

[0018] The technical solutions in the embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings.

[0019] The terms "first," "second," etc., in the description, claims, and accompanying drawings are used to distinguish different objects and not to describe a particular order. Furthermore, the term "comprising" and any variations thereof are intended to cover a non-exclusive inclusion. For example, a process, method, system, product, or apparatus that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to such processes, methods, products, or apparatus.

[0020] The AI-generated content front-end real-time recognition method based on bimodal verification provided in several embodiments of this specification can be executed by the AI-generated content front-end real-time recognition system based on bimodal verification provided in the embodiments of this invention.

[0021] Before this specification elaborates on the real-time front-end recognition method for AI-generated content based on bimodal verification in conjunction with one or more embodiments, it first introduces the application scenarios of this real-time front-end recognition method for AI-generated content based on bimodal verification.

[0022] Please see Figure 1 , Figure 1 This is a schematic diagram illustrating an application scenario of the AI-generated content front-end real-time recognition method based on bimodal verification provided in this embodiment of the invention. In this embodiment, the AI-generated content front-end real-time recognition system 100 based on bimodal verification may include a terminal 110, which may be a mobile phone, tablet computer, smart Bluetooth device, laptop computer, or personal computer (PC), etc. The terminal 110 is equipped with a front-end environment, meaning it has client applications, browsers, or other front-end frameworks capable of running on the user's device. For example, browsers such as Chrome, Firefox, and Safari allow users to access web applications; client applications such as mobile applications (iOS, Android) or desktop applications (Windows, macOS); and other front-end frameworks such as Electron-based desktop applications and React Native-based cross-platform applications.

[0023] In this embodiment, terminal 110 can be a mobile phone, tablet computer, smart Bluetooth device, laptop computer, or personal computer (PC), etc. Terminal 110 includes a central processing unit (CPU), a graphics processing unit (GPU), memory, storage devices, a network communication module, sensors, a display screen, a battery and power management module, etc. The central processing unit (CPU) can execute logical calculations, resource scheduling, and rendering optimization algorithms for front-end applications; the graphics processing unit (GPU) can accelerate front-end page rendering, especially the rendering of complex graphics, animations, and multimedia content; memory can store runtime data, resource files, user behavior data, and environmental awareness data for front-end applications; storage devices can store front-end application code, resource files, and cached data; the network communication module can communicate with the server to obtain dynamic resources and real-time environmental information; sensors can be used to perceive the environmental status of the user device in real time, such as network sensors for obtaining network bandwidth, latency, and connection status, performance sensors for obtaining CPU, GPU, and memory usage of the terminal 110, and screen sensors for obtaining screen resolution, brightness, and refresh rate; the display screen can be used to present the content of the front-end application; and the battery and power management module can provide power support for device operation and optimize energy consumption.

[0024] The terminal 110 in this embodiment can acquire user file information and user operation data based on the front-end environment running on the user device, and generate a behavior fingerprint based on the user operation data; perform edge density processing and word frequency jump detection processing on the user file information to obtain the content features of the user file information; determine the user's risk score data based on the content features and user operation data according to the preset dual-modal risk judgment engine; when the risk score data meets the first preset condition, determine the user file information as non-AI generated content, and add the behavior fingerprint to the metadata corresponding to the user file information to obtain the behavior fingerprint metadata.

[0025] It should be noted that, Figure 1 The schematic diagram of the AI-generated content front-end real-time recognition system 100 based on bimodal verification shown is merely an example. The AI-generated content front-end real-time recognition system and scenario based on bimodal verification described in this embodiment are for the purpose of more clearly illustrating the technical solutions of this embodiment and do not constitute a limitation on the technical solutions provided by this embodiment. As those skilled in the art will know, with the evolution of AI-generated content front-end real-time recognition systems based on bimodal verification and the emergence of new scenarios, the technical solutions provided by this embodiment are also applicable to similar technical problems.

[0026] Please see Figure 2 , Figure 2 This is a flowchart illustrating a real-time front-end recognition method for AI-generated content based on bimodal verification, provided in an embodiment of the present invention. This real-time front-end recognition method for AI-generated content based on bimodal verification can be... Figure 1 The AI-generated content front-end real-time recognition system 100 based on bimodal verification is executed. This AI-generated content front-end real-time recognition method based on bimodal verification may include at least the following steps: 200. Based on the front-end environment running on the user's device, obtain user file information and user operation data, and generate behavioral fingerprints based on the user operation data.

[0027] In this embodiment, the terminal 110 is equipped with a front-end environment, such as a browser (e.g., Chrome, Firefox, Safari, etc.) through which the user accesses web applications; a client application (e.g., mobile applications (iOS, Android) or desktop applications (Windows, macOS); or other front-end frameworks (e.g., Electron-based desktop applications, React Native-based cross-platform applications, etc.). User file information may include image information, text information, etc. User operation data refers to the user's behavioral data on user file information within the terminal's front-end environment. The behavioral fingerprint can be a hash value generated based on the user operation data.

[0028] In some embodiments, user operation data may include mouse movement path coordinates and timing, number of file edits, and operation time.

[0029] In some embodiments, please refer to Figure 3 , Figure 3 This is a schematic diagram of a process for generating behavioral fingerprints based on user operation data, provided in an embodiment of the present invention. Generating behavioral fingerprints based on user operation data includes: 2000. Obtain a random projection matrix, where any element in the random projection matrix is ​​randomly sampled from a standard normal distribution; 2010. Determine behavioral data vectors through user operation data; 2020. Multiply the behavioral data vector with the random projection matrix to obtain the low-dimensional projection vector; 2030. Based on the hash determination engine, determine the hash value of any dimension element in the low-dimensional projection vector; 2040. Generate behavioral fingerprints based on the hash values ​​of elements in any dimension of the low-dimensional projection vector.

[0030] In this embodiment, during the process of generating a behavior fingerprint based on user operation data, the terminal 110 can first obtain a d×n-dimensional random projection matrix, where any element in the random projection matrix is ​​randomly sampled from a standard normal distribution; then, the behavior data vector is determined through the user operation data, and the behavior data vector is multiplied by the random projection matrix to obtain a low-dimensional projection vector; for the low-dimensional projection vector, the behavior fingerprint can be calculated by calculating the hash value bit by bit based on the hash determination engine.

[0031] In some embodiments, the hash value of any dimension element in the low-dimensional projection vector is determined based on the hash determination engine, including: when the dimension element is less than a first preset value, the hash value corresponding to any dimension element is a first value; when the dimension element is not less than the first preset value, the hash value corresponding to any dimension element is a second value.

[0032] For example, for a low-dimensional projection vector p, calculate the elements of arbitrary dimensions p i When the hash value is obtained, the dimension element p i When less than 0, the dimension element p i The hash value is set to 0 when the dimension element p i When the dimension element p is not less than 0, i The hash value is set to 1. For example, for a low-dimensional projection vector p=[0.5,-0.3,1.2,...,-0.8], the hash value corresponding to the low-dimensional projection vector p is [1,0,1,...,0].

[0033] The embodiments in this specification can also convert the hash value corresponding to the low-dimensional projection vector p into a hexadecimal string, for example: A3F9C7D2. The embodiments in this specification can quickly compare the similarity of user behaviors using the generated behavioral fingerprint (such as a 16-bit string).

[0034] 210. Perform edge density processing and word frequency jump detection on the user file information to obtain the content features of the user file information.

[0035] In some embodiments, please refer to Figure 4 , Figure 4 This is a schematic flowchart illustrating the process of obtaining content features of user file information according to an embodiment of the present invention. The content features include image features and text features. Edge density processing and word frequency jump detection processing are performed on the user file information to obtain the content features of the user file information, including: 2100. Obtain image and text information from user files; 2110. Perform size processing on the image information to obtain an image with a preset size; 2120. Calculate the edge density of an image with a preset size, where the edge density is the image feature corresponding to the user file information; 2130. Obtain the abnormal punctuation rate corresponding to the text information; 2140. Perform word frequency jump detection processing on the text information to obtain word frequency jump detection data; 2150. Determine the contextual coherence score based on word frequency jump detection data; abnormal punctuation rate and contextual coherence score are text features corresponding to user file information.

[0036] In this embodiment, the terminal 110 can obtain an image with a preset size by performing size processing on the image in the image information, for example, setting the preset size to 100×100 pixels.

[0037] In some embodiments, calculating the edge density of an image with a preset size includes: obtaining the number of edge pixels and the total number of pixels of the image with the preset size, wherein the edge density is the ratio of the number of edge pixels to the total number of pixels.

[0038] In some embodiments, word frequency jump detection processing is performed on text information to obtain word frequency jump detection data, including: sliding the text information through a preset sliding window, determining the occurrence count of each word in each sliding window; determining the word frequency change of each word in adjacent sliding windows based on the occurrence count of each word; and obtaining word frequency jump detection data through the word frequency change.

[0039] In this embodiment, word frequency jump detection processing is performed on the text information. That is, the frequency of each word is determined in a sliding window, and the changes of these frequencies between different windows are observed. The specific steps include: counting word frequencies, that is, counting the number of times each word appears in each sliding window; calculating jump values, that is, comparing the word frequency change of the same word in adjacent sliding windows. If the word frequency change of a certain word in adjacent windows is large (i.e., "jump"), it is considered that the occurrence pattern of the word may be unnatural text.

[0040] For example, assuming the sliding window size is 3 and the text is "This is a test test text", the word frequency jump detection results include the first sliding window: "This is a test", with word frequencies of: {"this":1, "is":1, "a":1, "test":1}; the second sliding window: "A test test", with word frequencies of: {"a":1, "test":2}; and the third sliding window: "Test test text", with word frequencies of: {"test":2, "text":1}.

[0041] In this embodiment, the word frequency of the word "test" increases from 1 to 2 in the second window and then remains at 2 in the third window. This rapid change in word frequency may indicate anomalies in the text, such as excessive repetition or a lack of natural contextual transitions.

[0042] The embodiments in this specification can quantify the coherence of text based on the results of word frequency jump detection. The specific calculation method for the score can be as follows: define a jump threshold, setting a threshold for word frequency jumps (e.g., a word frequency change exceeding 10%); then count the number of jumps, that is, count the number of word frequency jumps exceeding the threshold in all sliding windows of the text; then calculate the coherence score, that is, calculate the coherence score based on the jump threshold and the number of jumps. The more jumps, the lower the coherence score; the fewer jumps, the higher the coherence score.

[0043] For example, in this embodiment of the specification, the jump threshold can be set to 10%. In the text "This is a test test text," the word frequency change rate of "test" from the first window to the second window is 100% (from 1 to 2), exceeding the 10% threshold. The word frequency change rate of "test" from the second window to the third window is 0% (remains unchanged). The word frequency change of "text" from the second window to the third window is 1. Although the rate of change cannot be calculated, it can be considered a significant change. Therefore, the word frequency jump of the word "test" is obvious, which may indicate an anomaly in the text. That is, the word frequency jump of "test" exceeds the threshold, so the coherence score is low. In contrast, the word frequency changes of natural text (such as "This is a test text used to detect contextual coherence") are more gradual, resulting in a higher coherence score.

[0044] 220. Based on a preset dual-modal risk assessment engine, determine the user's risk score data according to content characteristics and user operation data.

[0045] In some embodiments, based on a preset bimodal risk assessment engine, the user's risk score data is determined according to content features and user operation data, including: determining content risk score data based on content anomaly assessment rules in the bimodal risk assessment engine and content features; determining behavioral risk score data based on behavioral anomaly assessment rules in the bimodal risk assessment engine and behavior operation data; and obtaining the user's risk score data through the content risk score data and the behavioral risk score data.

[0046] In this embodiment, the bimodal risk assessment engine may include behavioral anomaly assessment rules and content anomaly assessment rules. Behavioral anomaly assessment rules are a set of rules used to determine the existence of abnormal behavior based on user operation data; that is, they identify potential abnormal behavior by analyzing user operation patterns, behavior frequency, operation paths, and other characteristics in the front-end system.

[0047] For example, in the rules for judging abnormal behavior, regarding whether the user has submitted a file without editing it, if the number of edits is 0, it can be judged as machine operation. Similarly, regarding the mouse path data detected by the terminal, Fourier transform analysis is used to determine its frequency characteristics. If the Fourier dominant frequency is greater than a preset value (e.g., 2Hz), it can be determined that there is a high-frequency repetitive pattern in the mouse path, which is more consistent with the characteristics of machine operation, thus indicating a high probability of machine operation. If the Fourier dominant frequency is not greater than the preset value, it indicates that the mouse path is relatively smooth and does not have obvious repetitive patterns, which is more consistent with the characteristics of human operation, thus indicating a low probability of machine operation.

[0048] For example, in the content anomaly detection rules for image files, if the terminal detects an edge density of less than 5%, it is highly likely that the file was generated by AI; in the content anomaly detection rules for text files, if the terminal detects an abnormal punctuation rate of >40%, it is highly likely that the file was generated by machine.

[0049] The embodiments of this specification can determine behavioral risk score data based on the behavioral anomaly judgment rules in the bimodal risk judgment engine and user operation data. The embodiments of this specification can also determine content risk score data based on the content anomaly judgment rules in the bimodal risk judgment engine and content characteristics, and then obtain the user's risk score data through the content risk score data and behavioral risk score data.

[0050] 230. When the risk scoring data meets the first preset condition, the user file information is determined to be non-AI generated content, and the behavioral fingerprint is added to the metadata corresponding to the user file information to obtain the behavioral fingerprint metadata.

[0051] In this embodiment, the first preset condition can be set as follows: the behavioral risk score data is less than the behavioral risk score preset value, the content risk score data is less than the content risk score preset value, or the sum of the content risk score data and the behavioral risk score data is less than the total score preset value.

[0052] In this embodiment, metadata refers to text information data, providing descriptive information about the data. Metadata may include information such as the data's source, format, creation time, author, and copyright. This embodiment can incorporate behavioral fingerprints into the metadata corresponding to user file information, thereby providing additional verification information for the file and helping to identify its authenticity and origin.

[0053] In some embodiments, the front-end real-time identification method for AI-generated content based on bimodal verification further includes: when the risk score data meets the second preset condition, determining the user file information as AI-generated content and intercepting the user file information.

[0054] In this embodiment, the second preset condition can be set as follows: the behavioral risk score data is not less than the behavioral risk score preset value, the content risk score data is not less than the content risk score preset value, and the sum of the content risk score data and the behavioral risk score data is not less than the total score preset value.

[0055] The embodiments in this specification employ a dual-modal verification method that combines user behavior and content. By comprehensively considering user operating habits and file content characteristics, the detection accuracy is improved. Machine-generated content is difficult to simultaneously forge human operating habits and natural content characteristics. Therefore, the embodiments in this specification can more effectively identify AI-generated content and reduce the false positive rate through dual-modal verification.

[0056] Moreover, the embodiments in this specification run on the front end, without relying on large AI models. Lightweight detection is achieved through a rule engine, which greatly reduces the consumption of computing resources and avoids network latency and data transmission costs caused by cloud computing.

[0057] Furthermore, the detection latency in this embodiment is reduced from 1200ms in cloud detection to 80ms, and the response time is less than 100ms, enabling rapid real-time interception. Eliminating network transmission and cloud queuing time eliminates noticeable lag when users upload images or text, greatly improving the user experience. This is particularly suitable for applications with high real-time requirements, such as content publishing on social media platforms.

[0058] Furthermore, the embodiments described in this specification can run locally, with user actions and file content processed locally without uploading to the cloud, effectively avoiding the risk of data leakage. Relying on native browser APIs, it complies with privacy regulations such as GDPR, fully protecting user privacy and data security.

[0059] Furthermore, the embodiments in this specification, through comprehensive analysis of user behavior and content characteristics, require attackers to simultaneously forge content features and simulate human operational behavior, significantly increasing the cost of forgery. Uploaders of AI-generated content typically exhibit characteristics of "high content quality + low operational complexity," while real users often employ multiple adjustments and random operational paths. The design of this invention can effectively identify this difference, thereby enhancing the ability to recognize AI-generated content and reducing the risk of being bypassed.

[0060] The embodiments in this specification solve the problems of single detection dimension, high computational cost, high privacy risk and poor real-time performance in the existing technology through innovative dual-modal verification technology. It provides an efficient, accurate, safe and real-time solution for the recognition of AI-generated content, which has broad application prospects and important practical significance.

[0061] The foregoing has described specific embodiments of this specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in a different order than that shown in the embodiments and may still achieve the desired result. Furthermore, the processes depicted in the drawings do not necessarily require the specific or sequential order shown to achieve the desired result. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0062] Please see Figure 5 , Figure 5 This is a schematic diagram of the structure of a front-end real-time recognition system for AI-generated content based on bimodal verification, provided as an embodiment of this specification.

[0063] like Figure 5 As shown, the AI-generated content front-end real-time recognition system based on bimodal verification may include at least the following: The data acquisition module 500 is used to acquire user file information and user operation data based on the front-end environment running on the user device, and generate behavior fingerprints based on the user operation data. The feature extraction module 510 is used to perform edge density processing and word frequency jump detection processing on user file information to obtain the content features of user file information. The risk assessment module 520 is used to determine the user's risk score data based on the content characteristics and user operation data, using a preset dual-modal risk assessment engine. The metadata tagging module 530 is used to determine that the user file information is not AI-generated content when the risk scoring data meets the first preset condition, and to add the behavior fingerprint to the metadata corresponding to the user file information to obtain the behavior fingerprint metadata.

[0064] In some embodiments, the data acquisition module 500 includes a fingerprint generation module, which is configured to: acquire a random projection matrix, wherein any element in the random projection matrix is ​​randomly sampled from a standard normal distribution; determine a behavior data vector through user operation data; multiply the behavior data vector with the random projection matrix to obtain a low-dimensional projection vector; determine the hash value of any dimension element in the low-dimensional projection vector based on a hash determination engine; and generate a behavior fingerprint based on the hash value of any dimension element in the low-dimensional projection vector.

[0065] In some embodiments, the data acquisition module 500 includes a hash determination module, which is used to: when the dimension element is less than a first preset value, the hash value corresponding to any dimension element is a first value; when the dimension element is not less than the first preset value, the hash value corresponding to any dimension element is a second value.

[0066] In some embodiments, the content features include image features and text features. The feature extraction module 510 includes a feature extraction submodule, which is used to: acquire image information and text information of user file information; perform size processing on the image in the image information to obtain an image with a preset size; calculate the edge density of the image with the preset size, where the edge density is the image feature corresponding to the user file information; acquire the abnormal punctuation rate corresponding to the text information; perform word frequency jump detection processing on the text information to obtain word frequency jump detection data; determine the context coherence score based on the word frequency jump detection data; and the abnormal punctuation rate and the context coherence score are the text features corresponding to the user file information.

[0067] In some embodiments, the feature extraction submodule includes an edge density module, which is used to: obtain the number of edge pixels and the total number of pixels of an image with a preset size, wherein the edge density is the ratio of the number of edge pixels to the total number of pixels.

[0068] In some embodiments, the feature extraction submodule includes a sliding window module, which is used to: slide the sliding window across the text information based on a preset sliding window, determine the occurrence count of each word in each sliding window; determine the word frequency change of each word in adjacent sliding windows based on the occurrence count of each word; and obtain word frequency jump detection data through the word frequency change.

[0069] In some embodiments, the risk assessment module 520 includes a scoring module, which is used to: determine content risk score data based on content characteristics according to the content anomaly assessment rules in the bimodal risk assessment engine; determine behavior risk score data based on user operation data according to the behavior anomaly assessment rules in the bimodal risk assessment engine; and obtain the user's risk score data through the content risk score data and the behavior risk score data.

[0070] In some embodiments, the AI-generated content front-end real-time recognition system based on bimodal verification further includes an interception module, which is used to: determine the user file information as AI-generated content when the risk score data meets the second preset condition, and intercept the user file information.

[0071] In some embodiments, user operation data includes mouse movement path coordinates and timing, number of file edits, and operation time.

[0072] This specification's embodiments, through innovative bimodal verification technology, solve the problems of single detection dimension, high computational cost, significant privacy risks, and poor real-time performance in existing technologies. It provides an efficient, accurate, secure, and real-time solution for the recognition of AI-generated content, with broad application prospects and important practical significance.

[0073] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to interchangeably. Each embodiment focuses on its differences from other embodiments. In particular, the embodiment of the AI-generated content front-end real-time recognition system based on bimodal verification is relatively simple in description because it is fundamentally similar to the embodiment of the AI-generated content front-end real-time recognition method based on bimodal verification; relevant parts can be referred to the description of the method embodiment.

[0074] This specification also provides a computer-readable storage medium storing instructions that, when executed on a computer or processor, cause the computer or processor to perform the above-described instructions. Figures 2-4 One or more steps in the illustrated embodiment. If the constituent modules of the above-described electronic device are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium.

[0075] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. This program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the methods described above. The aforementioned storage medium includes various media capable of storing program code, such as ROM, RAM, magnetic disks, or optical disks. Unless otherwise specified, the technical features of this embodiment and its implementation can be combined arbitrarily.

[0076] The embodiments described above are merely preferred embodiments of this specification and are not intended to limit the scope of this specification. Any modifications and improvements made by those skilled in the art to the technical solutions of this specification without departing from the spirit of this specification should fall within the protection scope defined by the claims of this specification.

Claims

1. A real-time front-end recognition method for AI-generated content based on bimodal verification, characterized in that, include: Based on the front-end environment running on the user's device, user file information and user operation data are obtained, and behavioral fingerprints are generated based on the user operation data. Edge density processing and word frequency jump detection processing are performed on the user file information to obtain the content features of the user file information; Based on a preset dual-modal risk assessment engine, the user's risk score data is determined according to the content features and the user operation data; When the risk score data meets the first preset condition, the user file information is determined to be non-AI generated content, and the behavior fingerprint is added to the metadata corresponding to the user file information to obtain the behavior fingerprint metadata. The step of generating a behavioral fingerprint based on the user operation data includes: Obtain a random projection matrix, wherein any element in the random projection matrix is ​​randomly sampled from a standard normal distribution; The behavior data vector is determined using the user operation data. The behavior data vector is multiplied by the random projection matrix to obtain a low-dimensional projection vector; Based on the hash determination engine, the hash value of any dimension element in the low-dimensional projection vector is determined; Generate a behavioral fingerprint based on the hash value of any dimension element in the low-dimensional projection vector; The content features include image features and text features. The process of performing edge density processing and word frequency jump detection on the user file information to obtain the content features of the user file information includes: Obtain the image and text information of the user file information; The image in the image information is resized to obtain an image with a preset size; Calculate the edge density of the image with a preset size, where the edge density is the image feature corresponding to the user file information; Obtain the abnormal punctuation rate corresponding to the text information; The text information is subjected to word frequency jump detection processing to obtain word frequency jump detection data; The contextual coherence score is determined based on word frequency jump detection data; the abnormal punctuation rate and the contextual coherence score are text features corresponding to the user file information.

2. The AI-generated content front-end real-time recognition method based on bimodal verification according to claim 1, characterized in that, The step of determining the hash value of any dimension element in the low-dimensional projection vector based on the hash determination engine includes: When the dimension element is less than a first preset value, the hash value corresponding to the dimension element is a first value; when the dimension element is not less than the first preset value, the hash value corresponding to the dimension element is a second value.

3. The AI-generated content front-end real-time recognition method based on bimodal verification according to claim 1, characterized in that, The calculation of the edge density of the image with the preset size includes: Obtain the number of edge pixels and the total number of pixels of the image with a preset size, wherein the edge density is the ratio of the number of edge pixels to the total number of pixels.

4. The AI-generated content front-end real-time recognition method based on bimodal verification according to claim 1, characterized in that, The step of performing word frequency jump detection processing on the text information to obtain word frequency jump detection data includes: Based on a preset sliding window, the sliding window is used to slide the text information, and the number of times each word appears in each sliding window is determined. The frequency change of each word in adjacent sliding windows is determined based on the number of times each word appears. The word frequency jump detection data is obtained by measuring the word frequency change.

5. The AI-generated content front-end real-time recognition method based on bimodal verification according to claim 1, characterized in that, The preset bimodal risk assessment engine determines the user's risk score data based on the content features and the user operation data, including: Based on the content anomaly determination rules in the dual-modal risk determination engine, content risk score data is determined according to the content characteristics; Based on the behavior anomaly judgment rules in the dual-modal risk judgment engine, behavior risk score data is determined according to the user operation data; The user's risk score data is obtained through the content risk score data and the behavior risk score data.

6. The AI-generated content front-end real-time recognition method based on bimodal verification according to claim 1, characterized in that, The method further includes: When the risk score data meets the second preset condition, the user file information is determined to be AI-generated content, and the user file information is intercepted.

7. The AI-generated content front-end real-time recognition method based on bimodal verification according to claim 1, characterized in that, The user operation data includes mouse movement path coordinates and timing, number of file edits, and operation time.

8. A real-time front-end recognition system for AI-generated content based on dual-modal verification, characterized in that, include: The data acquisition module is used to acquire user file information and user operation data based on the front-end environment running on the user device, and generate a behavior fingerprint based on the user operation data. The feature extraction module is used to perform edge density processing and word frequency jump detection processing on the user file information to obtain the content features of the user file information; The risk assessment module is used to determine the user's risk score data based on the content features and the user operation data, using a preset dual-modal risk assessment engine. The metadata tagging module is used to determine the user file information as non-AI generated content when the risk score data meets the first preset condition, and to add the behavior fingerprint to the metadata corresponding to the user file information to obtain behavior fingerprint metadata. The step of generating a behavioral fingerprint based on the user operation data includes: Obtain a random projection matrix, wherein any element in the random projection matrix is ​​randomly sampled from a standard normal distribution; The behavior data vector is determined using the user operation data. The behavior data vector is multiplied by the random projection matrix to obtain a low-dimensional projection vector; Based on the hash determination engine, the hash value of any dimension element in the low-dimensional projection vector is determined; Generate a behavioral fingerprint based on the hash value of any dimension element in the low-dimensional projection vector; The content features include image features and text features. The process of performing edge density processing and word frequency jump detection on the user file information to obtain the content features of the user file information includes: Obtain the image and text information of the user file information; The image in the image information is resized to obtain an image with a preset size; Calculate the edge density of the image with a preset size, where the edge density is the image feature corresponding to the user file information; Obtain the abnormal punctuation rate corresponding to the text information; The text information is subjected to word frequency jump detection processing to obtain word frequency jump detection data; The contextual coherence score is determined based on word frequency jump detection data; the abnormal punctuation rate and the contextual coherence score are text features corresponding to the user file information.

Images