System

Abstract

A system is provided.SOLUTION: A system including means for acquiring an advertisement or content, means for inputting the acquired advertisement or content to a AI model and evaluating reliability, means for specifying a fraudulent advertisement or content on the basis of an analysis result of the AI model, means for requesting deletion or correction to an advertiser who is confirmed to be fraudulent, and means for confirming a response situation of the advertiser and imposing a penalty in a case of non-compliance.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The technology of the present disclosure relates to a system. [Background technology]

[0002] Patent document 1 discloses a persona chatbot control method performed by at least one processor, the method including the steps of receiving a user utterance, adding the user utterance to a prompt including an instruction sentence related to a description of the chatbot character, encoding the prompt, and inputting the encoded prompt into a language model to generate a chatbot utterance in response to the user utterance. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Publication No. 2022-180282 Summary of the Invention [Problem to be solved by the invention]

[0004] Recently, there has been an increase in false advertisements and malicious deepfakes created using AI technology in web services. Such malicious content can cause harm to end users and damage the reputation of web service operators. The purpose of this invention is to solve these problems and build a healthy web environment. [Means for solving the problem]

[0005] The present invention is a system including a means for acquiring advertisements and content, a means for inputting the acquired advertisements and content into an AI model and evaluating their reliability, a means for identifying fraudulent advertisements and content based on the analysis results of the AI ​​model, a means for requesting deletion or correction from advertisers whose fraud has been confirmed, and a means for checking the advertiser's response status and imposing penalties if the advertiser does not comply. Also, the system includes a means for acquiring advertisements and content using an API or a scraping tool, and a means for requesting deletion or correction that includes a notification sending means.

[0006] "Advertisement" is visual or audio content created by a company or individual for the purpose of promoting a product, service, or information.

[0007] "Content" means digital data such as text, images, video, and audio, including information and entertainment provided on a web service.

[0008] "Acquisition method" refers to the technical methods and tools used to identify and download advertisements and content on the Web.

[0009] An "AI model" is a general term for an artificial intelligence algorithm that uses machine learning and deep learning based on large amounts of data to recognize patterns and make predictions and classifications.

[0010] "Measures to assess trustworthiness" refers to analytical methods used to determine whether advertisements or content are based on accurate and trustworthy information.

[0011] "Identification methods" are methods for identifying and classifying fraudulent advertisements and content based on the analysis results.

[0012] "Means for requesting removal or correction" refers to the methods and means of communication for requesting that an advertiser change or remove an advertisement or content that is deemed fraudulent.

[0013] "Notification Delivery Method" means a communication system or protocol for sending messages or alerts to Advertisers.

[0014] "Response confirmation means" refers to a technical means for tracking and confirming how an advertiser responds to a request for deletion or correction.

[0015] "Means of imposing penalties" refers to the imposition of sanctions, such as fines or usage restrictions, on advertisers when appropriate measures are not taken against fraudulent advertisements or content.

[0016] An "API" is an interface that allows different software applications to communicate and collaborate with each other.

[0017] "Scraping tool" refers to software or scripts used to automatically collect information from websites. [Brief explanation of the drawings]

[0018] [Figure 1] 1 is a conceptual diagram showing an example of the configuration of a data processing system according to a first embodiment. [Figure 2] 1 is a conceptual diagram showing an example of main functions of a data processing device and a smart device according to a first embodiment. [Figure 3] FIG. 10 is a conceptual diagram showing an example of the configuration of a data processing system according to a second embodiment. [Figure 4] FIG. 10 is a conceptual diagram showing an example of main functions of a data processing device and smart glasses according to a second embodiment. [Figure 5] FIG. 10 is a conceptual diagram showing an example of the configuration of a data processing system according to a third embodiment. [Figure 6] FIG. 11 is a conceptual diagram showing an example of main functions of a data processing device and a headset-type terminal according to a third embodiment. [Figure 7] FIG. 10 is a conceptual diagram showing an example of the configuration of a data processing system according to a fourth embodiment. [Figure 8]FIG. 10 is a conceptual diagram showing an example of main functions of a data processing device and a robot according to a fourth embodiment. [Figure 9] 1 shows an emotion map onto which multiple emotions are mapped. [Figure 10] 1 shows an emotion map onto which multiple emotions are mapped. [Figure 11] FIG. 3 is a sequence diagram showing a processing flow of the data processing system according to the first embodiment. [Figure 12] FIG. 10 is a sequence diagram showing the flow of processing in the data processing system in Application Example 1. [Figure 13] FIG. 10 is a sequence diagram showing the flow of processing in the data processing system according to the second embodiment when an emotion engine is combined. [Figure 14] FIG. 10 is a sequence diagram showing the flow of processing in the data processing system in Application Example 2 when an emotion engine is combined. DETAILED DESCRIPTION OF THE INVENTION

[0019] An example of an embodiment of a system according to the technology of the present disclosure will be described below with reference to the accompanying drawings.

[0020] First, the terms used in the following description will be explained.

[0021] In the following embodiments, a coded processor (hereinafter simply referred to as a "processor") may be a single arithmetic device or a combination of multiple arithmetic devices. Furthermore, a processor may be a single type of arithmetic device or a combination of multiple types of arithmetic devices. Examples of arithmetic devices include a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), a GPGPU (General-Purpose computing on Graphics Processing Units), and an APU (Accelerated Processing Unit).

[0022] In the following embodiments, a coded RAM (Random Access Memory) is a memory in which information is temporarily stored and is used as a working memory by a processor.

[0023] In the following embodiments, the coded storage is one or more non-volatile storage devices that store various programs, various parameters, etc. Examples of non-volatile storage devices include flash memory (SSD (Solid State Drive)), magnetic disks (e.g., hard disks), and magnetic tapes.

[0024] In the following embodiments, a communication I / F (Interface) with a symbol is an interface including a communication processor, an antenna, etc. The communication I / F controls communication between multiple computers. Examples of communication standards applied to the communication I / F include wireless communication standards including 5G (5th Generation Mobile Communication System), Wi-Fi (registered trademark), Bluetooth (registered trademark), etc.

[0025] In the following embodiments, "A and / or B" is synonymous with "at least one of A and B." In other words, "A and / or B" means that it may be only A, only B, or a combination of A and B. Furthermore, in this specification, the same concept as "A and / or B" is also applied when three or more things are expressed connected by "and / or."

[0026] [First embodiment]

[0027] FIG. 1 shows an example of the configuration of a data processing system 10 according to the first embodiment.

[0028] 1, a data processing system 10 includes a data processing device 12 and a smart device 14. An example of the data processing device 12 is a server.

[0029] The data processing device 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 is an example of a "computer" according to the technology of the present disclosure. The computer 22 includes a processor 28, a RAM 30, and a storage 32. The processor 28, the RAM 30, and the storage 32 are connected to a bus 34. The database 24 and the communication I / F 26 are also connected to the bus 34. The communication I / F 26 is connected to a network 54. Examples of the network 54 include a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[0030] The smart device 14 includes a computer 36, a reception device 38, an output device 40, a camera 42, and a communication I / F 44. The computer 36 includes a processor 46, a RAM 48, and a storage 50. The processor 46, the RAM 48, and the storage 50 are connected to a bus 52. The reception device 38, the output device 40, and the camera 42 are also connected to the bus 52.

[0031] The reception device 38 includes a touch panel 38A, a microphone 38B, and the like, and receives user input. The touch panel 38A detects contact with an indicator (for example, a pen or a finger) to receive user input by the touch of the indicator. The microphone 38B detects the user's voice to receive user input by voice. The control unit 46A transmits data indicating the user input received by the touch panel 38A and the microphone 38B to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the data indicating the user input.

[0032] The output device 40 includes a display 40A and a speaker 40B, and presents data to the user 20 by outputting the data in a form of expression that the user 20 can perceive (for example, audio and / or text). The display 40A displays visible information such as text and images in accordance with instructions from the processor 46. The speaker 40B outputs audio in accordance with instructions from the processor 46. The camera 42 is a compact digital camera equipped with an optical system including a lens, aperture, and shutter, and an imaging element such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor.

[0033] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 control the exchange of various information between the processor 46 and the processor 28 via the network 54.

[0034] FIG. 2 shows an example of the main functions of the data processing device 12 and the smart device 14.

[0035] 2, in the data processing device 12, a specific process is performed by the processor 28. A specific processing program 56 is stored in the storage 32. The specific processing program 56 is an example of a "program" according to the technology of the present disclosure. The processor 28 reads the specific processing program 56 from the storage 32 and executes the read specific processing program 56 on the RAM 30. The specific process is realized by the processor 28 operating as a specific processing unit 290 in accordance with the specific processing program 56 executed on the RAM 30.

[0036] The storage 32 stores a data generation model 58 and an emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[0037] In the smart device 14, the processor 46 performs the reception output process. The storage 50 stores a reception output program 60. The reception output program 60 is used in conjunction with the specific processing program 56 by the data processing system 10. The processor 46 reads the reception output program 60 from the storage 50 and executes the read reception output program 60 on the RAM 48. The reception output process is realized by the processor 46 operating as the control unit 46A in accordance with the reception output program 60 executed on the RAM 48.

[0038] Next, a description will be given of the specific processing performed by the specific processing unit 290 of the data processing device 12. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0039] ---

[0040] The "Fraudulent AI Checker" system of the present invention effectively monitors advertisements and content on web services and detects false information and deep fakes generated by the use of malicious AI technology. This system is implemented primarily in a server-based configuration.

[0041] The components of this system include the following means:

[0042] 1. How you get ads and content:

[0043] The server retrieves new ads and content from the target web service using an API or scraping tool.

[0044] 2. Input to the AI ​​model:

[0045] The server converts the acquired advertisements and content into the required format and inputs them into the AI ​​model.

[0046] 3. Reliability assessment measures:

[0047] AI models analyze data within ads and content to assess their trustworthiness, covering visual, audio, and textual elements.

[0048] 4. How to identify inappropriate content:

[0049] The server identifies fraudulent advertisements and content based on the analysis results of the AI ​​model.

[0050] 5. How to request deletion or correction:

[0051] The server provides a means for sending a notice to an advertiser where fraud has been confirmed, requesting deletion or correction.

[0052] 6. How to check compatibility:

[0053] The server monitors the response status of the advertiser and checks whether the response has been made within a specified time limit.

[0054] 7. Means of imposing penalties:

[0055] To provide a means for the server to impose penalties on advertisers if they do not take appropriate action.

[0056] ---

[0057] Program processing explanation

[0058] The server first retrieves newly posted ads and content from the target web service, specifically by downloading the data using an API or scraping tool and storing it in temporary storage.

[0059] The server then converts the stored ads and content into a more easily digestible format and feeds it into an AI model that analyzes the video, audio, and text information and compares it with public sources to assess its trustworthiness.

[0060] After the AI ​​model evaluates the reliability, the server receives the analysis results and identifies fraudulent ads and content. If the reliability score falls below a certain threshold, the content is deemed fraudulent and recorded in a dedicated database.

[0061] The server will send a notice to the advertiser requesting that the advertiser remove or correct any inappropriate content that has been recorded. If the advertiser does not respond within a specified timeframe, the server will impose penalties on the advertiser, such as automatically removing the advertisement or applying a fine.

[0062] ---

[0063] Specific examples

[0064] As an example, an embodiment in a video distribution service will be described.

[0065] 1. The server calls the API to retrieve new video ads on the distribution platform.

[0066] 2. The server converts the video and audio data it receives into an analyzable format for input into the AI ​​model.

[0067] 3. An AI model analyzes data in the video ad and detects that the person in the ad was created using deepfake technology.

[0068] 4. The server determines the ad is fraudulent and sends a notification to the advertiser requesting its removal.

[0069] 5. After the server notifies the user, it checks whether the advertisement has been deleted within the specified time limit, and if the user does not comply, it automatically deletes the advertisement and imposes a penalty.

[0070] Another example is its application in blog advertising.

[0071] 1. The server retrieves newly posted ads from the blog site.

[0072] 2. The server converts the ad into text data to be input into the AI ​​model.

[0073] 3. The AI ​​model analyzes all text in the ad to detect false product information and exaggerated claims of effectiveness.

[0074] 4. The server determines that the ad is a false advertisement and sends a notice to the advertiser requesting its removal.

[0075] 5. If the advertiser does not take appropriate action after the server notifies them, the advertisements will be automatically removed from the blog site and a fine will be imposed on the advertiser.

[0076] In this way, the "Malicious AI Checker" system based on this invention provides a concrete means for quickly detecting threats posed by malicious AI use and protecting end users and web service operators.

[0077] The processing flow will be explained below.

[0078] Step 1:

[0079] The server retrieves new ads and content from the target web service. Specifically, the server uses an API or a scraping tool to download data from the web service's endpoint.

[0080] Step 2:

[0081] The server stores the acquired ads and content in temporary storage, which is fast and reliable because it is a processing stage before the data is input into the AI ​​model.

[0082] Step 3:

[0083] The server converts the stored ads and content into a more easily processed format, for example extracting video frames, transcribing audio data, or removing unnecessary tags and formatting from text data.

[0084] Step 4:

[0085] The server inputs the converted data into the AI ​​model, which then makes an API call to the AI ​​model to send the data and begin analysis.

[0086] Step 5:

[0087] AI models analyze information within ads and content, such as facial recognition and movement analysis for video, speech recognition and sentiment analysis for audio, and natural language processing for text.

[0088] Step 6:

[0089] The AI ​​model sends the analysis results to a server, which include a reliability score, suspected fraud, and specific details.

[0090] Step 7:

[0091] The server receives the analysis results from the AI ​​model and identifies fraudulent ads and content. Specifically, if the reliability score falls below a set threshold, the ad or content is deemed fraudulent.

[0092] Step 8:

[0093] The server records any advertisements or content that it determines to be fraudulent in a dedicated database, which allows it to maintain a history and be used for subsequent processing.

[0094] Step 9:

[0095] The server will then send a notice to the advertiser requesting removal or correction of the identified fraudulent content, including the reason for the fraudulent content and instructions on how to correct or remove the content.

[0096] Step 10:

[0097] The server monitors the advertiser's response status, confirms whether the advertiser has responded within the specified time limit, and records the response history.

[0098] Step 11:

[0099] If the server does not comply with the advertiser's request, it will impose penalties, such as automatic deletion of advertisements or imposition of fines.

[0100] Step 12:

[0101] The server stores all processing results and response history as logs and periodically provides reports to the web service operator, including the number of fraudulent content cases detected and the advertiser's response status.

[0102] In this way, by performing specific actions sequentially at each step, the "Fraudulent AI Checker" system effectively detects fraudulent advertisements and content and takes measures against them.

[0103] Example 1

[0104] Next, a description will be given of Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0105] Currently, advertisements and content on web services sometimes contain false information or deepfakes generated using malicious AI technology. This can lead users to make decisions based on misinformation and believe in inappropriate content or advertisements. This fraudulent content is also a major problem for web service operators, as it undermines their credibility. A system that can quickly and effectively detect and respond to these issues is needed.

[0106] The specific processing by the specific processing unit 290 of the data processing device 12 in the first embodiment is realized by the following means.

[0107] In this invention, the server includes means for acquiring advertisements and content, means for converting the acquired advertisements and content into an appropriate data format, means for inputting the converted data into a generative AI model and evaluating its reliability, means for identifying fraudulent advertisements and content based on the analysis results of the generative AI model, means for requesting deletion or correction from advertisers whose fraud has been confirmed, and means for checking the advertiser's response status and imposing penalties if the advertiser does not comply. This makes it possible to quickly detect fraudulent advertisements and content on web services and take appropriate measures.

[0108] "Advertisements and content" refers to information and media published on web services that are intended to promote purchases or provide information to users.

[0109] "Means of acquisition" refers to APIs and scraping tools used to collect advertisements and content from web services.

[0110] "Means for converting into an appropriate data format" refers to the process for converting captured advertisements or content into an analyzable format, such as image, text, audio, or video data.

[0111] A "generative AI model" is a model that uses a pre-trained artificial intelligence algorithm to evaluate the reliability and accuracy of data based on the data provided.

[0112] "Measures to assess trustworthiness" refers to the process of using generative AI models to analyze advertising and content data to determine whether the content is accurate and whether it has been generated by malicious AI technology.

[0113] "Means for identifying fraudulent advertisements and content based on analysis results" refers to the process of using the reliability assessment results output from the generative AI model to identify whether advertisements and content are fraudulent.

[0114] "Means for requesting removal or correction" refers to a notification system that requests advertisers and content providers to remove or correct advertisements or content that are deemed fraudulent.

[0115] "Measures to verify response status" refers to a process for monitoring how advertisers and content providers respond to notifications and assessing whether appropriate action has been taken within the specified timeframe.

[0116] "Measures for imposing penalties" refers to a system that implements sanctions, such as automatic removal of advertisements or imposition of fines, if advertisers or content providers do not take action within a specified deadline.

[0117] MODE FOR CARRYING OUT THE INVENTION

[0118] The "Fraudulent AI Checker" system of this invention monitors advertisements and content on web services to detect false information and deep fakes generated using malicious AI technology. This system is implemented primarily in a server-based configuration.

[0119] System Configuration

[0120] The server is the main processor and includes the following elements:

[0121] 1. How you get ads and content:

[0122] The server retrieves data from the target web service using an API or a scraping tool. An example of an API is a REST API. Examples of scraping tools include Beautiful Soup and Scrapy.

[0123] 2. Data format conversion method:

[0124] The acquired advertisements and content are converted into the appropriate data format. For example, video files are extracted as images for each frame, and audio data is converted into text. OpenCV is used for image processing, and the Google Cloud Speech-to-Text API is used for speech-to-text conversion.

[0125] 3. Input to the AI ​​model:

[0126] The server then inputs the converted data into a generative AI model, which is built using TensorFlow or PyTorch, and the model is prompted with the prompt, "Please rate the reliability of this data."

[0127] 4. Reliability assessment measures:

[0128] The generative AI model analyzes the data and assesses its reliability, specifically assessing whether the video data was generated using deepfake technology and whether the text data contains exaggeration or false information.

[0129] 5. How to identify inappropriate content:

[0130] The server receives the analysis results from the AI ​​model and identifies fraudulent ads and content based on the reliability score. If fraud is determined, the results are recorded in a dedicated database.

[0131] 6. How to request deletion or correction:

[0132] The server sends notifications to advertisers who have been found to be fraudulent, requesting them to remove or correct the content. Notifications are sent via email systems or web service notification engines.

[0133] 7. How to check for compliance:

[0134] The server monitors the advertiser's response and checks whether the appropriate action was taken within the specified time period by calling the API again and checking the data.

[0135] 8. Means of imposing penalties:

[0136] If the advertiser does not respond within the specified time frame, the server will automatically remove the ads and content and impose fines or access restrictions as necessary.

[0137] Specific examples

[0138] Examples of applications for video streaming services include:

[0139] 1. The server retrieves a new video list using the video streaming service's API.

[0140] 2. To convert the video captured by the server into an analyzable format, the video file is cut out frame by frame and the audio data is converted into text.

[0141] 3. The server inputs the converted data into the deepfake detection AI model, entering the prompt "Please rate the reliability of this data."

[0142] 4. The AI ​​model analyzes the video frames and detects that they are deepfakes.

[0143] 5. The server determines that the video is fraudulent and records it in the database.

[0144] 6. The server sends a removal request notification to the advertiser.

[0145] 7. The server checks whether the video has been deleted within the specified time limit.

[0146] 8. If the server does not respond, the video will be automatically deleted and penalties will be applied.

[0147] In this way, the "Fraudulent AI Checker" system based on this invention provides a concrete means to quickly detect the threat of false information and deep fakes caused by fraudulent AI technology and protect users and web service operators.

[0148] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0149] Step 1:

[0150] Acquiring advertisements and content

[0151] The server retrieves ads and content from the target web service. Specifically, it collects data using an API or scraping tool. The retrieved data is then stored in temporary storage. The input is the web service's API endpoint or scraping tool settings, and the output is the retrieved ad and content data. This ensures that the latest ads and content are available within the system.

[0152] Step 2:

[0153] Data format conversion

[0154] The server converts the advertisements and content it acquires into an appropriate data format. Specifically, for video files, images are extracted frame by frame, and audio data is converted into text. The text information is used as is. The input is the acquired raw data, and the output is data converted into a format that can be analyzed by the generative AI model. This prepares data suitable for the AI ​​model.

[0155] Step 3:

[0156] Input to the AI ​​model

[0157] The server inputs the converted data into a generative AI model, which is built using TensorFlow or PyTorch. The model receives a prompt, "Please rate the reliability of this data." The input is the converted data and the prompt, and the output is the result of the AI ​​model's reliability assessment. This starts the process of assessing the reliability of the data.

[0158] Step 4:

[0159] Reliability assessment

[0160] The generative AI model analyzes the provided data and assesses its reliability. Specifically, it assesses whether deepfake technology has been used in video data and whether false or exaggerated information is included in text data. The input is a set of data to the AI ​​model, and the output is a reliability score or analysis results. This prepares the model for the next step based on the analysis results.

[0161] Step 5:

[0162] Identifying fraudulent content

[0163] The server receives the analysis results of the AI ​​model and identifies fraudulent ads and content based on the reliability score. A threshold score is set for what is deemed fraudulent, and content below this threshold is identified as fraudulent. The input is the reliability score and analysis results, and the output is a list of fraudulent content. This allows specific fraudulent content to be identified and addressed.

[0164] Step 6:

[0165] Requests for removal or correction

[0166] The server sends a notification to the advertiser where the fraud has been confirmed, requesting removal or correction. The notification is sent via an email system or a web service notification engine. The input is a list of fraudulent content and the advertiser's contact information, and the output is the notification sent. This allows the appropriate instructions to be given to the advertiser.

[0167] Step 7:

[0168] Check compatibility

[0169] The server monitors the advertiser's response and verifies whether appropriate action has been taken within the specified time limit. The server then calls the API again to retrieve data and verify whether the original inappropriate content has been corrected or deleted. The input is the re-retrieved data, and the output is the confirmation result of whether action has been taken. This makes it possible to understand the advertiser's response status.

[0170] Step 8:

[0171] Imposing a penalty

[0172] If the server does not take action within the specified deadline, it imposes penalties on the advertiser or content provider. Specifically, it automatically deletes the fraudulent ads and content, and imposes fines or access restrictions as necessary. The input is the response result and penalty conditions, and the output is the executed penalty. This allows inappropriate responses to be punished quickly, maintaining the reliability of the system.

[0173] (Application example 1)

[0174] Next, a description will be given of Application Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0175] In recent years, the use of malicious AI technology to create false information and deep fakes in advertisements and content has increased, increasing the likelihood of ordinary consumers being misled or harmed. There is a need for a means to quickly and effectively detect, correct, or remove such fraudulent advertisements and content. There is also a need for a system that can immediately warn viewers of the risk of fraudulent advertisements and urge them to be careful.

[0176] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 1 is realized by the following means.

[0177] In this invention, the server includes means for acquiring advertisements and content, means for inputting the acquired advertisements and content into an AI model and evaluating their reliability, means for identifying fraudulent advertisements and content based on the analysis results of the AI ​​model, means for requesting deletion or correction from advertisers whose fraud has been confirmed, means for checking the advertiser's response status and imposing penalties if the advertiser does not comply, means for automatically capturing advertisements viewed by users, means for analyzing the captured advertisements and evaluating their reliability, means for notifying users of a warning when a fraudulent advertisement is detected, and means for sending a warning to the advertiser as well. This improves the reliability of advertisements viewed by users and enables rapid detection and response to fraudulent advertisements.

[0178] "Means of obtaining advertisements and content" refers to the function of obtaining advertisements and content on web services using APIs or scraping tools.

[0179] "Means of inputting into an AI model and evaluating reliability" refers to a function that converts acquired advertisements and content into an appropriate format and inputs it into an AI model to evaluate its reliability.

[0180] "Means for identifying fraudulent advertisements and content" refers to a function that determines the reliability of advertisements and content based on the analysis results of an AI model and identifies those that contain fraudulent elements.

[0181] "Means to request removal or correction" is a function that sends a notification to the advertiser requesting removal or correction if fraudulent advertisements or content are confirmed.

[0182] "Means to check the advertiser's response status and impose penalties if they do not comply" is a function that checks whether the advertiser has responded appropriately to requests for deletion or correction, and imposes penalties if they do not comply.

[0183] "Means for automatically capturing advertisements viewed by users" refers to a function that automatically captures screenshots and URLs of advertisements viewed by users.

[0184] "Means for analyzing captured advertisements and evaluating their reliability" refers to a function that converts the captured advertisement data into an appropriate format and inputs it into an AI model to evaluate its reliability.

[0185] "Means for notifying users of a warning when fraudulent advertising is detected" is a function that displays a warning to users when fraudulent advertising is detected.

[0186] The "means for sending a warning notice to the advertiser" is a function that, when a fraudulent advertisement is detected, sends a warning notice to the advertiser that provided the advertisement.

[0187] The "Fraudulent AI Checker" system of this invention evaluates the reliability of advertisements and content viewed by users and detects false information and deep fakes. This system is implemented in a server-based configuration.

[0188] Components

[0189] 1. Server

[0190] 2. User device (smartphone)

[0191] 3. AI Model

[0192] Program structure and operation

[0193] 1. Acquiring advertisements and content

[0194] The server periodically retrieves new advertisements and content from the web service using an API or a scraping tool. The system also includes a function to automatically capture screenshots and URLs of advertisements viewed by the user's device. Smartphone screen capture APIs (such as Android's MediaProjection API and iOS's UIScreenshotService) are used for this purpose.

[0195] 2. Data conversion and input to AI models

[0196] The acquired advertisements and content are converted into an analyzable format by the server. For example, an image recognition API (such as Google Cloud Vision API) is used to convert the data into text or video data, which is then input into an AI model.

[0197] 3. Reliability evaluation

[0198] AI models (generative AI models built with TensorFlow, PyTorch, etc.) analyze ad and content data and evaluate their trustworthiness, with the aim of detecting deep fakes and false information.

[0199] 4. Identifying and warning fraudulent ads

[0200] Based on the analysis results of the AI ​​model, the server identifies fraudulent ads and content. If fraud is detected, the server displays a warning to the user. It also automatically sends a warning to the advertiser, requesting that the content be removed or corrected.

[0201] 5. Response confirmation and penalties

[0202] The server monitors whether the advertiser has responded appropriately to the deletion or correction request. If the request is not responded to within the specified time frame, the server will automatically delete the ad and impose a penalty on the advertiser. This notification is sent using services such as Firebase Cloud Messaging and SendGrid.

[0203] Specific examples

[0204] Application to video advertising

[0205] The server retrieves new video ads using the video distribution platform's API.

[0206] The acquired video advertisements are converted into a format that can be analyzed by the AI ​​model (video, audio).

[0207] AI model detects deepfakes in video ads.

[0208] The server determined the content to be fraudulent and sent a notification to the advertiser requesting its removal.

[0209] A warning message appears on the user's device stating, "This ad may be false."

[0210] If the advertiser does not comply, the server will automatically remove the video and impose a penalty.

[0211] Application to blog advertising

[0212] The server retrieves new ads from the blog site using a scraping tool.

[0213] Convert it into text data and input it into the AI ​​model.

[0214] AI models detect false product information and exaggerated claims.

[0215] The server determines the ad is fraudulent and notifies the advertiser to remove it.

[0216] A warning message appears on the user's device stating, "This ad may be false."

[0217] If advertisers do not comply, their ads will be automatically removed from blog sites and fines will be imposed.

[0218] Prompt Sentence Examples

[0219] "Enter your ad text and we'll check it for false or exaggerated information."

[0220] As described above, the system provides a concrete means to quickly assess the trustworthiness of advertisements and content, and protect users from false information and fraudulent advertisements.

[0221] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0222] Step 1: Getting ads and content

[0223] The server periodically obtains new advertisements and content from a web service using an API or scraping tool. The input is the web service's URL or API endpoint, and the output is the newly obtained advertisement or content data. Specifically, the server calls the API and receives JSON-formatted advertisement data as a response. The server also uses a scraping tool to extract advertisement information from HTML pages and temporarily store it.

[0224] Step 2: Capture the ads users see

[0225] The device automatically captures a screenshot or URL of the ad the user is viewing. The input is the ad screen being viewed, and the output is the captured screenshot or URL. Specifically, the device's screen capture API (for example, Android's MediaProjection API or iOS's UIScreenshotService) is used to capture an image of the ad being viewed.

[0226] Step 3: Data conversion

[0227] The server converts the acquired ad and content data into an analyzable format. The input is screenshots and ad data, and the output is text or video data. Specifically, it uses an image recognition API (such as Google Cloud Vision API) to extract text from ad screenshots and convert it into the required format. In the case of video data, it uses a video analysis algorithm to split it into analyzable frames.

[0228] Step 4: Input to the AI ​​model

[0229] The server inputs the converted data into an AI model. The input is text data or video data, and the output is a reliability evaluation result. Specifically, the data is input into a generative AI model built with TensorFlow or PyTorch, and reliability evaluation is performed. The AI ​​model performs analysis based on a pre-trained dataset.

[0230] Step 5: Reliability assessment

[0231] The server receives the analysis results of the AI ​​model and evaluates the reliability of the advertisements and content. The input is the reliability score obtained from the AI ​​model, and the output is the reliability evaluation result of the advertisement or content. Specifically, if the reliability score falls below a certain threshold, it is determined to be fraudulent and recorded in a dedicated database.

[0232] Step 6: Identifying and warning fraudulent ads

[0233] The server displays a warning to the user about ads or content that is determined to be fraudulent. The input is the reliability assessment result, and the output is a warning to the user. Specifically, it uses Firebase Cloud Messaging and the device's notification function to display a message on the user's device saying, "This ad may be false."

[0234] Step 7: Notify advertisers

[0235] The server automatically sends a warning to advertisers who have been found to be fraudulent, requesting them to remove or correct the content. The input is the reliability assessment result and the advertiser's contact information, and the output is the notification sent. Specifically, an email service such as SendGrid is used to send an email to the advertiser requesting removal.

[0236] Step 8: Action Verification and Penalties

[0237] The server monitors whether the advertiser has responded appropriately to requests for deletion or correction. The input is the advertiser's response status, and the output is the monitoring results and penalty processing. Specifically, if no response is made, the server will automatically delete the advertisement and take steps to impose a penalty on the advertiser.

[0238] This process strengthens users' defenses against untrustworthy ads and encourages advertisers to take action quickly.

[0239] Furthermore, an emotion engine that estimates the user's emotion may be combined. That is, the identification processing unit 290 may estimate the user's emotion using the emotion identification model 59 and perform identification processing using the user's emotion.

[0240] ---

[0241] By combining an emotion engine with the "Fraudulent AI Checker" system of the present invention, it is possible to monitor user reactions and further improve the reliability of advertisements and content based on the results. This system recognizes the emotions expressed by users while viewing or operating the content in real time, and reflects this data in the analysis of the AI ​​model, thereby more accurately evaluating the fraudulent nature of advertisements and content.

[0242] The components of this system include the following means:

[0243] 1. How you get ads and content:

[0244] The server retrieves new ads and content from the target web service using an API or scraping tool.

[0245] 2. Input to the AI ​​model:

[0246] The server converts the acquired advertisements and content into the required format and inputs them into the AI ​​model.

[0247] 3. Reliability assessment measures:

[0248] AI models analyze data within ads and content to assess their trustworthiness, including visual, audio, and textual elements.

[0249] 4. How to identify inappropriate content:

[0250] The server identifies fraudulent advertisements and content based on the analysis results of the AI ​​model.

[0251] 5. How to request deletion or correction:

[0252] The server provides a means for sending a notice to an advertiser where fraud has been confirmed, requesting deletion or correction.

[0253] 6. How to check compatibility:

[0254] The server monitors the response status of the advertiser and checks whether the response has been made within a specified time limit.

[0255] 7. Means of imposing penalties:

[0256] To provide a means for the server to impose penalties on advertisers if they do not take appropriate action.

[0257] 8. Emotion Engine:

[0258] The device monitors the user's reactions to what they are watching and doing in real time, collecting data. The emotion engine uses sensors such as cameras and microphones to analyze the user's facial expressions and tone of voice.

[0259] 9. Emotional data feedback methods:

[0260] The server feeds back the user's emotional data obtained from the emotion engine to the AI ​​model, allowing the user's actual reaction to be taken into account when evaluating trustworthiness.

[0261] ---

[0262] Program processing explanation

[0263] The server first retrieves newly posted ads and content from the target web service, specifically by downloading the data using an API or scraping tool and storing it in temporary storage.

[0264] The server then converts the stored ads and content into a more easily digestible format and feeds it into an AI model that analyzes the video, audio, and text information and compares it with public sources to assess its trustworthiness.

[0265] The server receives emotional data from the emotion engine on each device. When a user watches an advertisement or watches content and shows some kind of reaction to it, the emotion engine analyzes their facial expressions and tone of voice to determine their emotional state.

[0266] The AI ​​model takes emotional data into account when assessing the trustworthiness of ads and content. In addition to regular data analysis, the AI ​​model checks whether the user's emotional data indicates unnatural reactions, improving the accuracy of the assessment.

[0267] The server receives the analysis results from the AI ​​model and identifies fraudulent ads and content. If the reliability score falls below a certain threshold, the content is deemed fraudulent and recorded in a dedicated database.

[0268] The server will send a notice to the advertiser requesting that the advertiser remove or correct any inappropriate content that has been recorded. If the advertiser does not respond within a specified timeframe, the server will impose penalties on the advertiser, such as automatically removing the advertisement or applying a fine.

[0269] ---

[0270] Specific examples

[0271] As an example, an embodiment in a video distribution service will be described.

[0272] 1. The server retrieves new video ads on the distribution platform via API.

[0273] 2. The server converts the video and audio data it receives into an analyzable format for input into the AI ​​model.

[0274] 3. The user's device collects emotional data while watching the ad. The emotion engine uses facial recognition and voice analysis to analyze the user's emotional state in real time.

[0275] 4. The AI ​​model evaluates the authenticity of the video ad based on the data in the ad and the emotional data sent from the device, and if the user shows an abnormal emotional response, it determines that the ad is fraudulent.

[0276] 5. The server determines the ad is fraudulent and sends a notification to the advertiser requesting its removal.

[0277] 6. After the notification, the server will check whether the advertisement is removed within the specified time limit and monitor the compliance status. If the compliance is not met, the advertisement will be automatically removed and a fine will be imposed.

[0278] Another example is its application in blog advertising.

[0279] 1. The server retrieves newly posted ads from the blog site.

[0280] 2. The server converts the ad into text data to be input into the AI ​​model.

[0281] 3. The user's device collects emotional data while viewing the ad. The emotion engine analyzes the user's facial expressions and tone of voice to identify their emotional state.

[0282] 4. The AI ​​model analyzes all the text in the ad, takes into account emotional data, and if it finds a high likelihood of causing misunderstanding or distrust in the user, it will determine that the ad is false.

[0283] 5. The server determines that the ad is a false advertisement and sends a notice to the advertiser requesting its removal.

[0284] 6. If the advertiser does not take appropriate action after the server notifies them, the advertisements will be automatically removed from the blog site and a fine will be imposed on the advertiser.

[0285] In this way, the "Fraudulent AI Checker" system based on this invention utilizes an emotion engine to quickly and accurately detect threats posed by fraudulent AI use, providing a concrete means for protecting end users and web service operators.

[0286] The processing flow will be explained below.

[0287] Step 1:

[0288] The server obtains new ads and content from the target web service. Specifically, the server downloads the latest ad and content data from the web service endpoint using an API or scraping tool.

[0289] Step 2:

[0290] The server stores the retrieved advertisements and content in temporary storage, allowing the data to proceed to the next processing step without loss.

[0291] Step 3:

[0292] The server converts the stored ads and content into a more easily processed format, for example extracting video frames, transcribing audio data, or removing unnecessary tags and formatting from text data.

[0293] Step 4:

[0294] The server inputs the converted data into the AI ​​model, which then makes an API call to the AI ​​model to send the analyzed data, starting processing.

[0295] Step 5:

[0296] When a user's device starts displaying an advertisement or content, the device's emotion engine analyzes the user's facial expressions and voice in real time, for example, capturing the user's facial expressions with a camera and analyzing the tone of the voice with a microphone.

[0297] Step 6:

[0298] The device then sends the analyzed emotional data to the server, which includes the user's reactions to the displayed advertisements and content (such as surprise, disbelief, or anger).

[0299] Step 7:

[0300] AI models analyze data within ads and content to assess their trustworthiness, taking into account video, audio, and text information, as well as emotional data transmitted from the device.

[0301] Step 8:

[0302] The server receives the analysis results from the AI ​​model and identifies fraudulent ads and content. If the reliability score falls below a set threshold, the ad or content is deemed fraudulent.

[0303] Step 9:

[0304] The server records any ads or content that it determines to be fraudulent in a dedicated database, allowing it to maintain a history and detailed information about the fraudulent content.

[0305] Step 10:

[0306] The server will then send a notice to the advertiser requesting removal or correction of the fraudulent content, including the reason for the fraudulent content and how to correct it.

[0307] Step 11:

[0308] The server monitors the advertiser's response status, checks whether the advertiser has responded appropriately within the specified time period, and records the response details and their history.

[0309] Step 12:

[0310] The server will penalize advertisers if they fail to comply, automatically removing any abusive ads or content and notifying them of penalties if necessary.

[0311] Step 13:

[0312] The server stores all processing results and response history as logs and periodically provides reports to the web service operator, including the number of fraudulent content detected and the advertiser's response status.

[0313] ---

[0314] With this processing flow, the "Fraudulent AI Checker" system based on this invention utilizes an emotion engine to effectively identify fraudulent advertisements and content, providing a concrete means for protecting users and web service operators.

[0315] Example 2

[0316] Next, a description will be given of Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0317] The distribution of fraudulent information in modern advertising and content has become a problem. In particular, advertisements and content containing false information are rampant on the Internet, resulting in confusion for users and a loss of trust in web services. Therefore, it is necessary to quickly and accurately detect such fraudulent advertisements and content and take appropriate measures.

[0318] The identification process by the identification processing unit 290 of the data processing device 12 in Example 2 is realized by the following means. In this invention, the server includes means for acquiring advertisements and content, means for inputting the acquired advertisements and content into an AI model and evaluating their reliability, means for identifying fraudulent advertisements and content based on the analysis results of the AI ​​model, means for collecting emotional data of users viewing content on the user's terminal and feeding the data back to the AI ​​model, means for requesting deletion or correction from advertisers whose fraud has been confirmed, and means for checking the advertiser's response status and imposing penalties if the advertiser does not comply. This enables rapid and accurate detection of fraudulent advertisements and content and effective countermeasure requests to advertisers. The "advertisement and content acquisition means" refers to means used to acquire new advertisements and content from web services.

[0319] An "AI model" is an artificial intelligence algorithm used to analyze data and assess its reliability.

[0320] The "reliability evaluation means" is a means for evaluating the reliability of the advertisement or content based on the acquired advertisement or content.

[0321] "Means for identifying fraudulent advertisements and content" refers to means for identifying fraudulent advertisements and content based on the analysis results of an AI model.

[0322] The "emotion data collection means" is a means for collecting emotional data generated while a user is viewing advertisements or content.

[0323] "Emotional data feedback means" refers to a means for feeding back collected emotional data to an AI model.

[0324] "Means for requesting removal or correction" are means for requesting that advertisers whose advertisements or content have been found to be fraudulent remove or correct them.

[0325] The "compliance status confirmation means" is a means for monitoring the advertiser's compliance status.

[0326] The "penalty measure" is a measure for imposing a penalty on an advertiser if the advertiser does not take the specified action.

[0327] This invention uses a "Fraudulent AI Checker" system to monitor and evaluate the reliability of advertisements and content in real time. This system aims to more accurately evaluate the fraudulent nature of advertisements and content by recognizing the emotions expressed by users while viewing or operating the content in real time and reflecting this data in the analysis of an AI model.

[0328] System Configuration

[0329] How you get ads and content

[0330] The server accesses the target web service through an API or scraping tool to obtain new advertisements and content. For example, an API is used to obtain data from a video distribution platform, and a scraping tool is used for blog sites that contain text information.

[0331] Transforming data and inputting it into AI models

[0332] The server converts the acquired advertisements and content into an analyzable format. FFMpeg is used for video data, and Google's speech recognition system is used to convert audio data into text. Once converted, the data is formatted in JSON and input into the AI ​​model.

[0333] Reliability assessment tools

[0334] AI models analyze the video, audio, and text information in ads and content to assess their credibility, comparing it with statistical benchmarks and public databases.

[0335] Emotional data collection method

[0336] The device monitors the user's reactions in real time while watching and operating the device. Using the device's camera and microphone, the device analyzes the user's facial expressions and tone of voice to collect emotional data. For example, OpenCV and DeepFace are used to obtain facial expression data from camera footage, and the Google voice analysis system is used to obtain voice data.

[0337] Emotional Data Feedback Method

[0338] The server feeds back the user's emotional data obtained from the emotion engine to the AI ​​model, which then takes the user's emotional information into account when evaluating the reliability of the AI ​​model.

[0339] How to identify fraudulent ads and content

[0340] The server receives the analysis results from the AI ​​model and identifies fraudulent ads and content. Ads and content with a reliability score below a certain threshold are recorded as fraudulent in a dedicated database.

[0341] How to request deletion or correction

[0342] The server sends a notification to the advertiser that the fraud was detected, requesting them to remove or correct the fraud. For example, an email notification is sent using SendGrid or Mailchimp.

[0343] How to check the status of response

[0344] The server monitors the advertiser's response status and checks whether the response is made within the specified deadline. The advertiser's response status is tracked using a log monitoring system such as ElasticSearch or Kibana.

[0345] Penalty measures

[0346] The server imposes penalties if the advertiser does not take the action specified by the advertiser. It has the function to automatically delete advertisements and notify users of penalties.

[0347] Examples of concrete examples and prompts

[0348] Below are some examples and prompts:

[0349] Embodiment of video distribution service

[0350] example:

[0351] 1. The server retrieves new video ads on the distribution platform via API.

[0352] 2. The server converts the video and audio data it receives into an analyzable format for input into the AI ​​model.

[0353] 3. The user's device collects emotional data while watching the ad. The emotion engine uses facial recognition and voice analysis to analyze the user's emotional state in real time.

[0354] 4. The AI ​​model evaluates the authenticity of the video ad based on the data in the ad and the emotional data sent from the device, and if the user shows an abnormal emotional response, it determines that the ad is fraudulent.

[0355] 5. The server determines the ad is fraudulent and sends a notification to the advertiser requesting its removal.

[0356] 6. After the notification, the server will check whether the advertisement is removed within the specified time limit and monitor the compliance status. If the compliance is not met, the advertisement will be automatically removed and a fine will be imposed.

[0357] Prompt Sentence Examples

[0358] "Explain how the server retrieves new video ads from a web service and how it feeds them into the AI ​​model. Use the YouTube API as a concrete example."

[0359] "Please explain the mechanism by which the device collects emotional data from users while they are watching video ads. Please specify the sensors and analysis methods used."

[0360] Please explain in detail how your AI model uses sentiment data to evaluate the trustworthiness of ads, including the specific evaluation criteria and algorithms.

[0361] In this way, the "fraudulent AI checker" system based on this invention can achieve highly accurate monitoring and rapid response, protecting end users and web service operators.

[0362] The flow of the specific processing in the second embodiment will be described with reference to FIG. 13. Detailed Description of Processing Steps

[0363] Step 1:

[0364] The server accesses the target web service (e.g., a video streaming platform) and uses an API or scraping tool to obtain new advertising and content data.

[0365] Input: API access key for web service and scraping tool settings.

[0366] Specific behavior: The server sends a request to the specified endpoint (e.g., "https: / / example.api.com / getNewAds") and downloads the metadata and content data of the new ads.

[0367] Output: The acquired advertising and content data (e.g., video files, audio files, text data) is stored in temporary storage.

[0368] Step 2:

[0369] The server converts the acquired advertisements and content into an analyzable format and inputs it into the AI ​​model.

[0370] Input: The original data of the ads and content stored.

[0371] What it does: The server splits the video file into frames using FFMpeg and converts the audio to text, which is then formatted into JSON.

[0372] Output: Data is generated in a parsable format (JSON) that is then fed into the AI ​​model.

[0373] Step 3:

[0374] The device collects the user's reactions in real time while watching and operating the device, and analyzes emotional data.

[0375] Input: Facial and voice data of the user while watching ads and content.

[0376] Specific operation: Using the device's camera and microphone, facial expressions are analyzed using OpenCV and DeepFace, and voice tone is analyzed using the Google voice analysis system.

[0377] Output: Data representing the user's emotional state (e.g., happiness, surprise, anger, sadness, etc.) is generated and sent to the server in real time.

[0378] Step 4:

[0379] The AI ​​model assesses credibility based on data and emotional data within the ad or content.

[0380] Input: Ad / content data and user sentiment data converted into an analyzable format.

[0381] How it works: The AI ​​model analyzes video, audio, and text information, compares it to statistical benchmarks and public databases, and integrates user sentiment data to perform pattern matching and anomaly detection to detect fraud.

[0382] Output: The AI ​​model generates a credibility score for each ad or piece of content.

[0383] Step 5:

[0384] The server receives the analysis results from the AI ​​model and identifies fraudulent ads and content.

[0385] Input: Analysis results of the AI ​​model (confidence score and fraud detection information).

[0386] How it works: The server checks the reliability score and lists ads and content that fall below a certain threshold. It also records this fraudulent data in a dedicated database.

[0387] Output: A list of abusive ads and content will be generated.

[0388] Step 6:

[0389] The server sends a notice to the advertiser where fraud has been confirmed, requesting removal or correction.

[0390] Input: A list of ads and content that have been identified as fraudulent, along with details about them.

[0391] What happens next: The server uses a specialized notification system (e.g., SendGrid, Mailchimp) to send an email to the advertiser requesting removal or correction. The notification will include the specific details of the fraud and a deadline for action.

[0392] Output: A notification is sent to the advertiser.

[0393] Step 7:

[0394] The server monitors the advertiser's response status and checks whether the response is made within the specified time limit. If the appropriate response is not made, a penalty will be imposed.

[0395] Input: Advertiser response reports and log data.

[0396] Specific actions: Using log monitoring systems such as ElasticSearch and Kibana, we track the advertiser's response status. If the response is not made within the deadline, we will automatically remove the ads and apply fines.

[0397] Output: Notification of the action status report and any penalties applied, if applicable.

[0398] Through the above processing steps, the system enables rapid and accurate detection and response to fraudulent advertisements and content.

[0399] (Application example 2)

[0400] Next, a description will be given of Application Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0401] In recent years, with the increase in online advertising and content, the reliability of fraudulent advertising and content has become a problem. This problem increases the risk of users receiving misinformation, leading to a decline in trust in advertising and content providers. In such situations, there is a need for systems that can detect and respond to fraud with high accuracy and speed. However, current systems have few methods for evaluating reliability using emotional data, making it difficult to detect fraud that reflects users' actual reactions.

[0402] The identification process by the identification processing unit 290 of the data processing device 12 in Application Example 2 is realized by the following means. In this invention, the server includes: means for acquiring advertisements and content; means for inputting the acquired advertisements and content into an AI model to evaluate their reliability; means having an emotion engine for collecting user emotion data and evaluating the reliability of advertisements and content based on the collected emotion data; means for identifying fraudulent advertisements and content based on the analysis results of the AI ​​model; means for requesting deletion or correction from advertisers whose fraud has been confirmed; and means for checking the advertiser's response and imposing penalties if they do not comply. This makes it possible to utilize user emotion data to more accurately evaluate the fraudulence of advertisements and content and respond quickly.

[0403] "Means for obtaining advertisements and content" refers to the technical methods for accessing and collecting advertisements and content, specifically, the means for obtaining digital information using programs such as APIs and scraping tools.

[0404] "Means for inputting acquired advertisements and content into an AI model to evaluate its reliability" refers to a series of processes for appropriately processing acquired digital information and inputting it into an AI model to analyze the reliability of that information.

[0405] "Emotion engine for collecting user emotional data" refers to technology that uses sensors and analytical algorithms to collect a user's facial expressions, tone of voice, and other data in real time to identify the user's emotional state.

[0406] "Means for evaluating the reliability of advertisements and content based on collected emotional data" refers to a mechanism that incorporates user emotional data into the analysis results and uses that data to more accurately evaluate the reliability of advertisements and content.

[0407] "Means for identifying fraudulent advertisements and content based on the analysis results of an AI model" refers to a method for automatically identifying fraudulent or suspicious advertisements and content based on data analyzed by an AI model.

[0408] "Means for requesting removal or correction from advertisers where fraud has been confirmed" refers to a method of sending a notice to an advertiser requesting the removal or correction of content when fraud is determined to have occurred.

[0409] "Measures to check the advertiser's response and impose penalties if they do not comply" refers to methods for monitoring whether the advertiser has responded to the notice and, if they do not respond, implementing penalties such as fines or automatic removal of ads.

[0410] MODE FOR CARRYING OUT THE INVENTION

[0411] An embodiment of the present invention is a system that uses user emotion data to evaluate the reliability of advertisements and content. In this system, a server mainly plays the following roles.

[0412] System Configuration

[0413] 1. How you get ads and content:

[0414] The server uses APIs or scraping tools to retrieve newly posted ads and content, giving you access to the latest digital information.

[0415] 2. Input to the AI ​​model:

[0416] The server converts the acquired advertisements and content into an analyzable format and inputs it into an AI model, which analyzes video, audio, and text information to evaluate its reliability.

[0417] 3. Emotion Engine:

[0418] The user's device is equipped with an emotion engine that uses a camera and microphone to analyze the user's facial expressions and tone of voice in real time to collect emotion data, using image processing libraries such as OpenCV and dlib.

[0419] 4. Reliability assessment measures:

[0420] The server evaluates the reliability of advertisements and content based on AI models and user emotional data. For example, if the user's emotional response is unnatural, the server determines that the advertisement or content is fraudulent.

[0421] 5. How to identify inappropriate content:

[0422] The server identifies fraudulent advertisements and content based on the analysis results of the AI ​​model, making it possible to quickly identify unreliable digital information.

[0423] 6. How to request deletion or correction:

[0424] The server sends notifications to advertisers who have been found to be fraudulent, requesting them to remove or correct the content. Notifications are efficiently handled through an automated system.

[0425] 7. How to check for compliance:

[0426] The server monitors the advertiser's response and checks whether the response is made within the specified time limit, thereby ensuring that appropriate action is taken.

[0427] 8. Means of imposing penalties:

[0428] The server will impose penalties on advertisers if they do not take appropriate action, such as automatically removing ads or applying fines.

[0429] Specific examples

[0430] Specifically, an embodiment in a video distribution service will be described.

[0431] 1. The server retrieves new video ads on the distribution platform via API.

[0432] 2. The server converts the video and audio data it receives into an analyzable format for input into the AI ​​model.

[0433] 3. The user's device collects emotional data while watching the ad. The emotion engine uses facial recognition and voice analysis to analyze the user's emotional state in real time.

[0434] 4. The AI ​​model evaluates the authenticity of the video ad based on the data in the ad and the emotional data sent from the device, and if the user shows an abnormal emotional response, it determines that the ad is fraudulent.

[0435] 5. The server determines the ad is fraudulent and sends a notification to the advertiser requesting its removal.

[0436] 6. After the notification, the server will check whether the advertisement is removed within the specified time limit and monitor the compliance status. If the compliance is not met, the advertisement will be automatically removed and a fine will be imposed.

[0437] Example prompts to input to the generative AI model

[0438] Below is an example of a prompt sentence to input to the generative AI model.

[0439] I want to create an "Emotional Ad Checker" app that evaluates the reliability of ad content. It analyzes the user's facial expressions and voice in real time using a camera and microphone, and evaluates the reliability of the ad. The process proceeds based on the following conditions:

[0440] 1. Get advertising data from the API.

[0441] 2. Capture user data with camera and microphone.

[0442] 3. Identify emotions from facial expressions and voice.

[0443] 4. Emotional data is sent to the server to evaluate the credibility of the advertisement.

[0444] The above is a specific embodiment for carrying out the present invention.

[0445] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[0446] Step 1:

[0447] The server retrieves advertisements and content. Digital information is collected from web services using APIs or scraping tools and temporarily stored. The input is data obtained via APIs or scraping tools, and the output is advertisement and content data stored in temporary storage.

[0448] Step 2:

[0449] The advertisements and content acquired by the server are converted into an analyzable format. Video, audio, and text information are arranged into the required format. The input is the data acquired in step 1, and the output is data converted into a format suitable for the AI ​​model. Specific operations include splitting video data into frames and sampling audio data.

[0450] Step 3:

[0451] The server inputs the converted ad and content data into the AI ​​model and evaluates its reliability. The AI ​​model analyzes the data and generates a reliability score. The input is the data prepared in step 2, and the output is the reliability score and analysis results. Specifically, image recognition and text analysis are performed using a neural network.

[0452] Step 4:

[0453] The device uses an emotion engine to collect real-time facial and vocal data of users watching advertisements and content. The input is video and audio data captured by a camera and microphone, and the output is analyzed emotional data. Specific operations include facial recognition and voice tone analysis.

[0454] Step 5:

[0455] The device sends the collected emotion data to the server. The input is the analyzed emotion data, and the output is the data transmission to the server. The specific operation includes executing an HTTP request to send the emotion data to the server.

[0456] Step 6:

[0457] The server combines the analysis results of the AI ​​model with the emotional data to evaluate the trustworthiness of the advertisement or content. The input is the trustworthiness score from step 3 and the emotional data from step 5, and the output is the final trustworthiness evaluation result. Specific operations include an algorithm that integrates the emotional data and the trustworthiness score and reassess the fraud risk.

[0458] Step 7:

[0459] The server identifies fraudulent ads and content based on the results of the reliability evaluation. The input is the reliability evaluation result from step 6, and the output is a list of ads and content that are determined to be fraudulent. Specific operations include a process of comparing the evaluation result with a threshold and listing those that meet the criteria for determining fraud.

[0460] Step 8:

[0461] The server sends notifications to advertisers that have been found to be fraudulent, requesting their removal or correction. The input is a list of ads or content that have been found to be fraudulent, and the output is the notifications sent to the advertisers. Specific operations include sending notifications using an automated email system.

[0462] Step 9:

[0463] The server checks the advertiser's response status and monitors whether the response is made within the specified time limit. The input is the advertiser's response status data, and the output is a log file of the response status. Specific operations include tracking the advertiser's response using a real-time monitoring system.

[0464] Step 10:

[0465] The server imposes penalties when advertisers do not take appropriate action. The input is log data on the response status, and the output is the penalty action taken. Specific actions include automatic removal of ads and application of fines.

[0466] The specific processing unit 290 transmits the result of the specific processing to the smart device 14. In the smart device 14, the control unit 46A causes the output device 40 to output the result of the specific processing. The microphone 38B acquires audio indicating a user input regarding the result of the specific processing. The control unit 46A transmits audio data indicating the user input acquired by the microphone 38B to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the audio data.

[0467] The data generation model 58 is a so-called generative AI (Artificial Intelligence). An example of the data generation model 58 is ChatGPT (Internet Search<URL: https: / / openai.com / blog / chatgpt> ), Gemini (Internet search <url: https: gemini.google.com ?hl="ja">) and other generation AIs. The data generation model 58 is obtained by performing deep learning on a neural network. A prompt including an instruction is input to the data generation model 58, and inference data such as voice data indicating voice, text data indicating text, and image data indicating an image is also input. The data generation model 58 performs inference on the input inference data in accordance with the instruction indicated by the prompt, and outputs the inference result in a data format such as voice data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[0468] In the above embodiment, an example in which the specific process is performed by the data processing device 12 has been given, but the technology of the present disclosure is not limited to this, and the specific process may be performed by the smart device 14.

[0469] [Second embodiment]

[0470] FIG. 3 shows an example of the configuration of a data processing system 210 according to the second embodiment.

[0471] 3, the data processing system 210 includes the data processing device 12 and smart glasses 214. An example of the data processing device 12 is a server.

[0472] The data processing device 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 is an example of a "computer" according to the technology of the present disclosure. The computer 22 includes a processor 28, a RAM 30, and a storage 32. The processor 28, the RAM 30, and the storage 32 are connected to a bus 34. The database 24 and the communication I / F 26 are also connected to the bus 34. The communication I / F 26 is connected to a network 54. Examples of the network 54 include a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[0473] The smart glasses 214 include a computer 36, a microphone 238, a speaker 240, a camera 42, and a communication I / F 44. The computer 36 includes a processor 46, a RAM 48, and a storage 50. The processor 46, the RAM 48, and the storage 50 are connected to a bus 52. The microphone 238, the speaker 240, and the camera 42 are also connected to the bus 52.

[0474] The microphone 238 receives instructions and the like from the user 20 by receiving voice uttered by the user 20. The microphone 238 captures the voice uttered by the user 20, converts the captured voice into audio data, and outputs it to the processor 46. The speaker 240 outputs audio in accordance with instructions from the processor 46.

[0475] Camera 42 is a small digital camera equipped with an optical system including a lens, aperture, and shutter, and an imaging element such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, and captures images of the surroundings of user 20 (for example, an imaging range defined by an angle of view equivalent to the field of vision of a typical healthy person).

[0476] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 are responsible for the exchange of various information between the processor 46 and the processor 28 via the network 54. The exchange of various information between the processor 46 and the processor 28 using the communication I / Fs 44 and 26 is carried out in a secure state.

[0477] Fig. 4 shows an example of the main functions of the data processing device 12 and the smart glasses 214. As shown in Fig. 4, in the data processing device 12, a specific process is performed by the processor 28. A specific process program 56 is stored in the storage 32.

[0478] The specific processing program 56 is an example of a "program" according to the technology of the present disclosure. The processor 28 reads the specific processing program 56 from the storage 32 and executes the read specific processing program 56 on the RAM 30. The specific processing is realized by the processor 28 operating as a specific processing unit 290 in accordance with the specific processing program 56 executed on the RAM 30.

[0479] The storage 32 stores a data generation model 58 and an emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[0480] In the smart glasses 214, the reception output process is performed by the processor 46. A reception output program 60 is stored in the storage 50. The processor 46 reads the reception output program 60 from the storage 50 and executes the read reception output program 60 on the RAM 48. The reception output process is realized by the processor 46 operating as the control unit 46A in accordance with the reception output program 60 executed on the RAM 48.

[0481] Next, a description will be given of the identification process performed by the identification processing unit 290 of the data processing device 12. In the following description, the data processing device 12 will be referred to as the "server" and the smart glasses 214 will be referred to as the "terminal."

[0482] ---

[0483] The "Fraudulent AI Checker" system of the present invention effectively monitors advertisements and content on web services and detects false information and deep fakes generated by the use of malicious AI technology. This system is implemented primarily in a server-based configuration.

[0484] The components of this system include the following means:

[0485] 1. How you get ads and content:

[0486] The server retrieves new ads and content from the target web service using an API or scraping tool.

[0487] 2. Input to the AI ​​model:

[0488] The server converts the acquired advertisements and content into the required format and inputs them into the AI ​​model.

[0489] 3. Reliability assessment measures:

[0490] AI models analyze data within ads and content to assess their trustworthiness, covering visual, audio, and textual elements.

[0491] 4. How to identify inappropriate content:

[0492] The server identifies fraudulent advertisements and content based on the analysis results of the AI ​​model.

[0493] 5. How to request deletion or correction:

[0494] The server provides a means for sending a notice to an advertiser where fraud has been confirmed, requesting deletion or correction.

[0495] 6. How to check compatibility:

[0496] The server monitors the response status of the advertiser and checks whether the response has been made within a specified time limit.

[0497] 7. Means of imposing penalties:

[0498] To provide a means for the server to impose penalties on advertisers if they do not take appropriate action.

[0499] ---

[0500] Program processing explanation

[0501] The server first retrieves newly posted ads and content from the target web service, specifically by downloading the data using an API or scraping tool and storing it in temporary storage.

[0502] The server then converts the stored ads and content into a more easily digestible format and feeds it into an AI model that analyzes the video, audio, and text information and compares it with public sources to assess its trustworthiness.

[0503] After the AI ​​model evaluates the reliability, the server receives the analysis results and identifies fraudulent ads and content. If the reliability score falls below a certain threshold, the content is deemed fraudulent and recorded in a dedicated database.

[0504] The server will send a notice to the advertiser requesting that the advertiser remove or correct any inappropriate content that has been recorded. If the advertiser does not respond within a specified timeframe, the server will impose penalties on the advertiser, such as automatically removing the advertisement or applying a fine.

[0505] ---

[0506] Specific examples

[0507] As an example, an embodiment in a video distribution service will be described.

[0508] 1. The server calls the API to retrieve new video ads on the distribution platform.

[0509] 2. The server converts the video and audio data it receives into an analyzable format for input into the AI ​​model.

[0510] 3. An AI model analyzes data in the video ad and detects that the person in the ad was created using deepfake technology.

[0511] 4. The server determines the ad is fraudulent and sends a notification to the advertiser requesting its removal.

[0512] 5. After the server notifies the user, it checks whether the advertisement has been deleted within the specified time limit, and if the user does not comply, it automatically deletes the advertisement and imposes a penalty.

[0513] Another example is its application in blog advertising.

[0514] 1. The server retrieves newly posted ads from the blog site.

[0515] 2. The server converts the ad into text data to be input into the AI ​​model.

[0516] 3. The AI ​​model analyzes all text in the ad to detect false product information and exaggerated claims of effectiveness.

[0517] 4. The server determines that the ad is a false advertisement and sends a notice to the advertiser requesting its removal.

[0518] 5. If the advertiser does not take appropriate action after the server notifies them, the advertisements will be automatically removed from the blog site and a fine will be imposed on the advertiser.

[0519] In this way, the "Malicious AI Checker" system based on this invention provides a concrete means for quickly detecting threats posed by malicious AI use and protecting end users and web service operators.

[0520] The processing flow will be explained below.

[0521] Step 1:

[0522] The server retrieves new ads and content from the target web service. Specifically, the server uses an API or a scraping tool to download data from the web service's endpoint.

[0523] Step 2:

[0524] The server stores the acquired ads and content in temporary storage, which is fast and reliable because it is a processing stage before the data is input into the AI ​​model.

[0525] Step 3:

[0526] The server converts the stored ads and content into a more easily processed format, for example extracting video frames, transcribing audio data, or removing unnecessary tags and formatting from text data.

[0527] Step 4:

[0528] The server inputs the converted data into the AI ​​model, which then makes an API call to the AI ​​model to send the data and begin analysis.

[0529] Step 5:

[0530] AI models analyze information within ads and content, such as facial recognition and movement analysis for video, speech recognition and sentiment analysis for audio, and natural language processing for text.

[0531] Step 6:

[0532] The AI ​​model sends the analysis results to a server, which include a reliability score, suspected fraud, and specific details.

[0533] Step 7:

[0534] The server receives the analysis results from the AI ​​model and identifies fraudulent ads and content. Specifically, if the reliability score falls below a set threshold, the ad or content is deemed fraudulent.

[0535] Step 8:

[0536] The server records any advertisements or content that it determines to be fraudulent in a dedicated database, which allows it to maintain a history and be used for subsequent processing.

[0537] Step 9:

[0538] The server will then send a notice to the advertiser requesting removal or correction of the identified fraudulent content, including the reason for the fraudulent content and instructions on how to correct or remove the content.

[0539] Step 10:

[0540] The server monitors the advertiser's response status, confirms whether the advertiser has responded within the specified time limit, and records the response history.

[0541] Step 11:

[0542] If the server does not comply with the advertiser's request, it will impose penalties, such as automatic deletion of advertisements or imposition of fines.

[0543] Step 12:

[0544] The server stores all processing results and response history as logs and periodically provides reports to the web service operator, including the number of fraudulent content cases detected and the advertiser's response status.

[0545] In this way, by performing specific actions sequentially at each step, the "Fraudulent AI Checker" system effectively detects fraudulent advertisements and content and takes measures against them.

[0546] Example 1

[0547] Next, a description will be given of Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal."

[0548] Currently, advertisements and content on web services sometimes contain false information or deepfakes generated using malicious AI technology. This can lead users to make decisions based on misinformation and believe in inappropriate content or advertisements. This fraudulent content is also a major problem for web service operators, as it undermines their credibility. A system that can quickly and effectively detect and respond to these issues is needed.

[0549] The specific processing by the specific processing unit 290 of the data processing device 12 in the first embodiment is realized by the following means.

[0550] In this invention, the server includes means for acquiring advertisements and content, means for converting the acquired advertisements and content into an appropriate data format, means for inputting the converted data into a generative AI model and evaluating its reliability, means for identifying fraudulent advertisements and content based on the analysis results of the generative AI model, means for requesting deletion or correction from advertisers whose fraud has been confirmed, and means for checking the advertiser's response status and imposing penalties if the advertiser does not comply. This makes it possible to quickly detect fraudulent advertisements and content on web services and take appropriate measures.

[0551] "Advertisements and content" refers to information and media published on web services that are intended to promote purchases or provide information to users.

[0552] "Means of acquisition" refers to APIs and scraping tools used to collect advertisements and content from web services.

[0553] "Means for converting into an appropriate data format" refers to the process for converting captured advertisements or content into an analyzable format, such as image, text, audio, or video data.

[0554] A "generative AI model" is a model that uses a pre-trained artificial intelligence algorithm to evaluate the reliability and accuracy of data based on the data provided.

[0555] "Measures to assess trustworthiness" refers to the process of using generative AI models to analyze advertising and content data to determine whether the content is accurate and whether it has been generated by malicious AI technology.

[0556] "Means for identifying fraudulent advertisements and content based on analysis results" refers to the process of using the reliability assessment results output from the generative AI model to identify whether advertisements and content are fraudulent.

[0557] "Means for requesting removal or correction" refers to a notification system that requests advertisers and content providers to remove or correct advertisements or content that are deemed fraudulent.

[0558] "Measures to verify response status" refers to a process for monitoring how advertisers and content providers respond to notifications and assessing whether appropriate action has been taken within the specified timeframe.

[0559] "Measures for imposing penalties" refers to a system that implements sanctions, such as automatic removal of advertisements or imposition of fines, if advertisers or content providers do not take action within a specified deadline.

[0560] MODE FOR CARRYING OUT THE INVENTION

[0561] The "Fraudulent AI Checker" system of this invention monitors advertisements and content on web services to detect false information and deep fakes generated using malicious AI technology. This system is implemented primarily in a server-based configuration.

[0562] System Configuration

[0563] The server is the main processor and includes the following elements:

[0564] 1. How you get ads and content:

[0565] The server retrieves data from the target web service using an API or a scraping tool. An example of an API is a REST API. Examples of scraping tools include Beautiful Soup and Scrapy.

[0566] 2. Data format conversion method:

[0567] The acquired advertisements and content are converted into the appropriate data format. For example, video files are extracted as images for each frame, and audio data is converted into text. OpenCV is used for image processing, and the Google Cloud Speech-to-Text API is used for speech-to-text conversion.

[0568] 3. Input to the AI ​​model:

[0569] The server then inputs the converted data into a generative AI model, which is built using TensorFlow or PyTorch, and the model is prompted with the prompt, "Please rate the reliability of this data."

[0570] 4. Reliability assessment measures:

[0571] The generative AI model analyzes the data and assesses its reliability, specifically assessing whether the video data was generated using deepfake technology and whether the text data contains exaggeration or false information.

[0572] 5. How to identify inappropriate content:

[0573] The server receives the analysis results from the AI ​​model and identifies fraudulent ads and content based on the reliability score. If fraud is determined, the results are recorded in a dedicated database.

[0574] 6. How to request deletion or correction:

[0575] The server sends notifications to advertisers who have been found to be fraudulent, requesting them to remove or correct the content. Notifications are sent via email systems or web service notification engines.

[0576] 7. How to check for compliance:

[0577] The server monitors the advertiser's response and checks whether the appropriate action was taken within the specified time period by calling the API again and checking the data.

[0578] 8. Means of imposing penalties:

[0579] If the advertiser does not respond within the specified time frame, the server will automatically remove the ads and content and impose fines or access restrictions as necessary.

[0580] Specific examples

[0581] Examples of applications for video streaming services include:

[0582] 1. The server retrieves a new video list using the video streaming service's API.

[0583] 2. To convert the video captured by the server into an analyzable format, the video file is cut out frame by frame and the audio data is converted into text.

[0584] 3. The server inputs the converted data into the deepfake detection AI model, entering the prompt "Please rate the reliability of this data."

[0585] 4. The AI ​​model analyzes the video frames and detects that they are deepfakes.

[0586] 5. The server determines that the video is fraudulent and records it in the database.

[0587] 6. The server sends a removal request notification to the advertiser.

[0588] 7. The server checks whether the video has been deleted within the specified time limit.

[0589] 8. If the server does not respond, the video will be automatically deleted and penalties will be applied.

[0590] In this way, the "Fraudulent AI Checker" system based on this invention provides a concrete means to quickly detect the threat of false information and deep fakes caused by fraudulent AI technology and protect users and web service operators.

[0591] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0592] Step 1:

[0593] Acquiring advertisements and content

[0594] The server retrieves ads and content from the target web service. Specifically, it collects data using an API or scraping tool. The retrieved data is then stored in temporary storage. The input is the web service's API endpoint or scraping tool settings, and the output is the retrieved ad and content data. This ensures that the latest ads and content are available within the system.

[0595] Step 2:

[0596] Data format conversion

[0597] The server converts the advertisements and content it acquires into an appropriate data format. Specifically, for video files, images are extracted frame by frame, and audio data is converted into text. The text information is used as is. The input is the acquired raw data, and the output is data converted into a format that can be analyzed by the generative AI model. This prepares data suitable for the AI ​​model.

[0598] Step 3:

[0599] Input to the AI ​​model

[0600] The server inputs the converted data into a generative AI model, which is built using TensorFlow or PyTorch. The model receives a prompt, "Please rate the reliability of this data." The input is the converted data and the prompt, and the output is the result of the AI ​​model's reliability assessment. This starts the process of assessing the reliability of the data.

[0601] Step 4:

[0602] Reliability assessment

[0603] The generative AI model analyzes the provided data and assesses its reliability. Specifically, it assesses whether deepfake technology has been used in video data and whether false or exaggerated information is included in text data. The input is a set of data to the AI ​​model, and the output is a reliability score or analysis results. This prepares the model for the next step based on the analysis results.

[0604] Step 5:

[0605] Identifying fraudulent content

[0606] The server receives the analysis results of the AI ​​model and identifies fraudulent ads and content based on the reliability score. A threshold score is set for what is deemed fraudulent, and content below this threshold is identified as fraudulent. The input is the reliability score and analysis results, and the output is a list of fraudulent content. This allows specific fraudulent content to be identified and addressed.

[0607] Step 6:

[0608] Requests for removal or correction

[0609] The server sends a notification to the advertiser where the fraud has been confirmed, requesting removal or correction. The notification is sent via an email system or a web service notification engine. The input is a list of fraudulent content and the advertiser's contact information, and the output is the notification sent. This allows the appropriate instructions to be given to the advertiser.

[0610] Step 7:

[0611] Check compatibility

[0612] The server monitors the advertiser's response and verifies whether appropriate action has been taken within the specified time limit. The server then calls the API again to retrieve data and verify whether the original inappropriate content has been corrected or deleted. The input is the re-retrieved data, and the output is the confirmation result of whether action has been taken. This makes it possible to understand the advertiser's response status.

[0613] Step 8:

[0614] Imposing a penalty

[0615] If the server does not take action within the specified deadline, it imposes penalties on the advertiser or content provider. Specifically, it automatically deletes the fraudulent ads and content, and imposes fines or access restrictions as necessary. The input is the response result and penalty conditions, and the output is the executed penalty. This allows inappropriate responses to be punished quickly, maintaining the reliability of the system.

[0616] (Application example 1)

[0617] Next, a description will be given of Application Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal."

[0618] In recent years, the use of malicious AI technology to create false information and deep fakes in advertisements and content has increased, increasing the likelihood of ordinary consumers being misled or harmed. There is a need for a means to quickly and effectively detect, correct, or remove such fraudulent advertisements and content. There is also a need for a system that can immediately warn viewers of the risk of fraudulent advertisements and urge them to be careful.

[0619] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 1 is realized by the following means.

[0620] In this invention, the server includes means for acquiring advertisements and content, means for inputting the acquired advertisements and content into an AI model and evaluating their reliability, means for identifying fraudulent advertisements and content based on the analysis results of the AI ​​model, means for requesting deletion or correction from advertisers whose fraud has been confirmed, means for checking the advertiser's response status and imposing penalties if the advertiser does not comply, means for automatically capturing advertisements viewed by users, means for analyzing the captured advertisements and evaluating their reliability, means for notifying users of a warning when a fraudulent advertisement is detected, and means for sending a warning to the advertiser as well. This improves the reliability of advertisements viewed by users and enables rapid detection and response to fraudulent advertisements.

[0621] "Means of obtaining advertisements and content" refers to the function of obtaining advertisements and content on web services using APIs or scraping tools.

[0622] "Means of inputting into an AI model and evaluating reliability" refers to a function that converts acquired advertisements and content into an appropriate format and inputs it into an AI model to evaluate its reliability.

[0623] "Means for identifying fraudulent advertisements and content" refers to a function that determines the reliability of advertisements and content based on the analysis results of an AI model and identifies those that contain fraudulent elements.

[0624] "Means to request removal or correction" is a function that sends a notification to the advertiser requesting removal or correction if fraudulent advertisements or content are confirmed.

[0625] "Means to check the advertiser's response status and impose penalties if they do not comply" is a function that checks whether the advertiser has responded appropriately to requests for deletion or correction, and imposes penalties if they do not comply.

[0626] "Means for automatically capturing advertisements viewed by users" refers to a function that automatically captures screenshots and URLs of advertisements viewed by users.

[0627] "Means for analyzing captured advertisements and evaluating their reliability" refers to a function that converts the captured advertisement data into an appropriate format and inputs it into an AI model to evaluate its reliability.

[0628] "Means for notifying users of a warning when fraudulent advertising is detected" is a function that displays a warning to users when fraudulent advertising is detected.

[0629] The "means for sending a warning notice to the advertiser" is a function that, when a fraudulent advertisement is detected, sends a warning notice to the advertiser that provided the advertisement.

[0630] The "Fraudulent AI Checker" system of this invention evaluates the reliability of advertisements and content viewed by users and detects false information and deep fakes. This system is implemented in a server-based configuration.

[0631] Components

[0632] 1. Server

[0633] 2. User device (smartphone)

[0634] 3. AI Model

[0635] Program structure and operation

[0636] 1. Acquiring advertisements and content

[0637] The server periodically retrieves new advertisements and content from the web service using an API or a scraping tool. The system also includes a function to automatically capture screenshots and URLs of advertisements viewed by the user's device. Smartphone screen capture APIs (such as Android's MediaProjection API and iOS's UIScreenshotService) are used for this purpose.

[0638] 2. Data conversion and input to AI models

[0639] The acquired advertisements and content are converted into an analyzable format by the server. For example, an image recognition API (such as Google Cloud Vision API) is used to convert the data into text or video data, which is then input into an AI model.

[0640] 3. Reliability evaluation

[0641] AI models (generative AI models built with TensorFlow, PyTorch, etc.) analyze ad and content data and evaluate their trustworthiness, with the aim of detecting deep fakes and false information.

[0642] 4. Identifying and warning fraudulent ads

[0643] Based on the analysis results of the AI ​​model, the server identifies fraudulent ads and content. If fraud is detected, the server displays a warning to the user. It also automatically sends a warning to the advertiser, requesting that the content be removed or corrected.

[0644] 5. Response confirmation and penalties

[0645] The server monitors whether the advertiser has responded appropriately to the deletion or correction request. If the request is not responded to within the specified time frame, the server will automatically delete the ad and impose a penalty on the advertiser. This notification is sent using services such as Firebase Cloud Messaging and SendGrid.

[0646] Specific examples

[0647] Application to video advertising

[0648] The server retrieves new video ads using the video distribution platform's API.

[0649] The acquired video advertisements are converted into a format that can be analyzed by the AI ​​model (video, audio).

[0650] AI model detects deepfakes in video ads.

[0651] The server determined the content to be fraudulent and sent a notification to the advertiser requesting its removal.

[0652] A warning message appears on the user's device stating, "This ad may be false."

[0653] If the advertiser does not comply, the server will automatically remove the video and impose a penalty.

[0654] Application to blog advertising

[0655] The server retrieves new ads from the blog site using a scraping tool.

[0656] Convert it into text data and input it into the AI ​​model.

[0657] AI models detect false product information and exaggerated claims.

[0658] The server determines the ad is fraudulent and notifies the advertiser to remove it.

[0659] A warning message appears on the user's device stating, "This ad may be false."

[0660] If advertisers do not comply, their ads will be automatically removed from blog sites and fines will be imposed.

[0661] Prompt Sentence Examples

[0662] "Enter your ad text and we'll check it for false or exaggerated information."

[0663] As described above, the system provides a concrete means to quickly assess the trustworthiness of advertisements and content, and protect users from false information and fraudulent advertisements.

[0664] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0665] Step 1: Getting ads and content

[0666] The server periodically obtains new advertisements and content from a web service using an API or scraping tool. The input is the web service's URL or API endpoint, and the output is the newly obtained advertisement or content data. Specifically, the server calls the API and receives JSON-formatted advertisement data as a response. The server also uses a scraping tool to extract advertisement information from HTML pages and temporarily store it.

[0667] Step 2: Capture the ads users see

[0668] The device automatically captures a screenshot or URL of the ad the user is viewing. The input is the ad screen being viewed, and the output is the captured screenshot or URL. Specifically, the device's screen capture API (for example, Android's MediaProjection API or iOS's UIScreenshotService) is used to capture an image of the ad being viewed.

[0669] Step 3: Data conversion

[0670] The server converts the acquired ad and content data into an analyzable format. The input is screenshots and ad data, and the output is text or video data. Specifically, it uses an image recognition API (such as Google Cloud Vision API) to extract text from ad screenshots and convert it into the required format. In the case of video data, it uses a video analysis algorithm to split it into analyzable frames.

[0671] Step 4: Input to the AI ​​model

[0672] The server inputs the converted data into an AI model. The input is text data or video data, and the output is a reliability evaluation result. Specifically, the data is input into a generative AI model built with TensorFlow or PyTorch, and reliability evaluation is performed. The AI ​​model performs analysis based on a pre-trained dataset.

[0673] Step 5: Reliability assessment

[0674] The server receives the analysis results of the AI ​​model and evaluates the reliability of the advertisements and content. The input is the reliability score obtained from the AI ​​model, and the output is the reliability evaluation result of the advertisement or content. Specifically, if the reliability score falls below a certain threshold, it is determined to be fraudulent and recorded in a dedicated database.

[0675] Step 6: Identifying and warning fraudulent ads

[0676] The server displays a warning to the user about ads or content that is determined to be fraudulent. The input is the reliability assessment result, and the output is a warning to the user. Specifically, it uses Firebase Cloud Messaging and the device's notification function to display a message on the user's device saying, "This ad may be false."

[0677] Step 7: Notify advertisers

[0678] The server automatically sends a warning to advertisers who have been found to be fraudulent, requesting them to remove or correct the content. The input is the reliability assessment result and the advertiser's contact information, and the output is the notification sent. Specifically, an email service such as SendGrid is used to send an email to the advertiser requesting removal.

[0679] Step 8: Action Verification and Penalties

[0680] The server monitors whether the advertiser has responded appropriately to requests for deletion or correction. The input is the advertiser's response status, and the output is the monitoring results and penalty processing. Specifically, if no response is made, the server will automatically delete the advertisement and take steps to impose a penalty on the advertiser.

[0681] This process strengthens users' defenses against untrustworthy ads and encourages advertisers to take action quickly.

[0682] Furthermore, an emotion engine that estimates the user's emotion may be further combined. That is, the identification processing unit 290 may estimate the user's emotion using the emotion identification model 59, and perform identification processing using the user's emotion.

[0683] ---

[0684] By combining an emotion engine with the "Fraudulent AI Checker" system of the present invention, it is possible to monitor user reactions and further improve the reliability of advertisements and content based on the results. This system recognizes the emotions expressed by users while viewing or operating the content in real time, and reflects this data in the analysis of the AI ​​model, thereby more accurately evaluating the fraudulent nature of advertisements and content.

[0685] The components of this system include the following means:

[0686] 1. How you get ads and content:

[0687] The server retrieves new ads and content from the target web service using an API or scraping tool.

[0688] 2. Input to the AI ​​model:

[0689] The server converts the acquired advertisements and content into the required format and inputs them into the AI ​​model.

[0690] 3. Reliability assessment measures:

[0691] AI models analyze data within ads and content to assess their trustworthiness, including visual, audio, and textual elements.

[0692] 4. How to identify inappropriate content:

[0693] The server identifies fraudulent advertisements and content based on the analysis results of the AI ​​model.

[0694] 5. How to request deletion or correction:

[0695] The server provides a means for sending a notice to an advertiser where fraud has been confirmed, requesting deletion or correction.

[0696] 6. How to check compatibility:

[0697] The server monitors the response status of the advertiser and checks whether the response has been made within a specified time limit.

[0698] 7. Means of imposing penalties:

[0699] To provide a means for the server to impose penalties on advertisers if they do not take appropriate action.

[0700] 8. Emotion Engine:

[0701] The device monitors the user's reactions to what they are watching and doing in real time, collecting data. The emotion engine uses sensors such as cameras and microphones to analyze the user's facial expressions and tone of voice.

[0702] 9. Emotional data feedback methods:

[0703] The server feeds back the user's emotional data obtained from the emotion engine to the AI ​​model, allowing the user's actual reaction to be taken into account when evaluating trustworthiness.

[0704] ---

[0705] Program processing explanation

[0706] The server first retrieves newly posted ads and content from the target web service, specifically by downloading the data using an API or scraping tool and storing it in temporary storage.

[0707] The server then converts the stored ads and content into a more easily digestible format and feeds it into an AI model that analyzes the video, audio, and text information and compares it with public sources to assess its trustworthiness.

[0708] The server receives emotional data from the emotion engine on each device. When a user watches an advertisement or watches content and shows some kind of reaction to it, the emotion engine analyzes their facial expressions and tone of voice to determine their emotional state.

[0709] The AI ​​model takes emotional data into account when assessing the trustworthiness of ads and content. In addition to regular data analysis, the AI ​​model checks whether the user's emotional data indicates unnatural reactions, improving the accuracy of the assessment.

[0710] The server receives the analysis results from the AI ​​model and identifies fraudulent ads and content. If the reliability score falls below a certain threshold, the content is deemed fraudulent and recorded in a dedicated database.

[0711] The server will send a notice to the advertiser requesting that the advertiser remove or correct any inappropriate content that has been recorded. If the advertiser does not respond within a specified timeframe, the server will impose penalties on the advertiser, such as automatically removing the advertisement or applying a fine.

[0712] ---

[0713] Specific examples

[0714] As an example, an embodiment in a video distribution service will be described.

[0715] 1. The server retrieves new video ads on the distribution platform via API.

[0716] 2. The server converts the video and audio data it receives into an analyzable format for input into the AI ​​model.

[0717] 3. The user's device collects emotional data while watching the ad. The emotion engine uses facial recognition and voice analysis to analyze the user's emotional state in real time.

[0718] 4. The AI ​​model evaluates the authenticity of the video ad based on the data in the ad and the emotional data sent from the device, and if the user shows an abnormal emotional response, it determines that the ad is fraudulent.

[0719] 5. The server determines the ad is fraudulent and sends a notification to the advertiser requesting its removal.

[0720] 6. After the notification, the server will check whether the advertisement is removed within the specified time limit and monitor the compliance status. If the compliance is not met, the advertisement will be automatically removed and a fine will be imposed.

[0721] Another example is its application in blog advertising.

[0722] 1. The server retrieves newly posted ads from the blog site.

[0723] 2. The server converts the ad into text data to be input into the AI ​​model.

[0724] 3. The user's device collects emotional data while viewing the ad. The emotion engine analyzes the user's facial expressions and tone of voice to identify their emotional state.

[0725] 4. The AI ​​model analyzes all the text in the ad, takes into account emotional data, and if it finds a high likelihood of causing misunderstanding or distrust in the user, it will determine that the ad is false.

[0726] 5. The server determines that the ad is a false advertisement and sends a notice to the advertiser requesting its removal.

[0727] 6. If the advertiser does not take appropriate action after the server notifies them, the advertisements will be automatically removed from the blog site and a fine will be imposed on the advertiser.

[0728] In this way, the "Fraudulent AI Checker" system based on this invention utilizes an emotion engine to quickly and accurately detect threats posed by fraudulent AI use, providing a concrete means for protecting end users and web service operators.

[0729] The processing flow will be explained below.

[0730] Step 1:

[0731] The server obtains new ads and content from the target web service. Specifically, the server downloads the latest ad and content data from the web service endpoint using an API or scraping tool.

[0732] Step 2:

[0733] The server stores the retrieved advertisements and content in temporary storage, allowing the data to proceed to the next processing step without loss.

[0734] Step 3:

[0735] The server converts the stored ads and content into a more easily processed format, for example extracting video frames, transcribing audio data, or removing unnecessary tags and formatting from text data.

[0736] Step 4:

[0737] The server inputs the converted data into the AI ​​model, which then makes an API call to the AI ​​model to send the analyzed data, starting processing.

[0738] Step 5:

[0739] When a user's device starts displaying an advertisement or content, the device's emotion engine analyzes the user's facial expressions and voice in real time, for example, capturing the user's facial expressions with a camera and analyzing the tone of the voice with a microphone.

[0740] Step 6:

[0741] The device then sends the analyzed emotional data to the server, which includes the user's reactions to the displayed advertisements and content (such as surprise, disbelief, or anger).

[0742] Step 7:

[0743] AI models analyze data within ads and content to assess their trustworthiness, taking into account video, audio, and text information, as well as emotional data transmitted from the device.

[0744] Step 8:

[0745] The server receives the analysis results from the AI ​​model and identifies fraudulent ads and content. If the reliability score falls below a set threshold, the ad or content is deemed fraudulent.

[0746] Step 9:

[0747] The server records any ads or content that it determines to be fraudulent in a dedicated database, allowing it to maintain a history and detailed information about the fraudulent content.

[0748] Step 10:

[0749] The server will then send a notice to the advertiser requesting removal or correction of the fraudulent content, including the reason for the fraudulent content and how to correct it.

[0750] Step 11:

[0751] The server monitors the advertiser's response status, checks whether the advertiser has responded appropriately within the specified time period, and records the response details and their history.

[0752] Step 12:

[0753] The server will penalize advertisers if they fail to comply, automatically removing any abusive ads or content and notifying them of penalties if necessary.

[0754] Step 13:

[0755] The server stores all processing results and response history as logs and periodically provides reports to the web service operator, including the number of fraudulent content detected and the advertiser's response status.

[0756] ---

[0757] With this processing flow, the "Fraudulent AI Checker" system based on this invention utilizes an emotion engine to effectively identify fraudulent advertisements and content, providing a concrete means for protecting users and web service operators.

[0758] Example 2

[0759] Next, a description will be given of Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal."

[0760] The distribution of fraudulent information in modern advertising and content has become a problem. In particular, advertisements and content containing false information are rampant on the Internet, resulting in confusion for users and a loss of trust in web services. Therefore, it is necessary to quickly and accurately detect such fraudulent advertisements and content and take appropriate measures.

[0761] The identification process by the identification processing unit 290 of the data processing device 12 in Example 2 is realized by the following means. In this invention, the server includes means for acquiring advertisements and content, means for inputting the acquired advertisements and content into an AI model and evaluating their reliability, means for identifying fraudulent advertisements and content based on the analysis results of the AI ​​model, means for collecting emotional data of users viewing content on the user's terminal and feeding the data back to the AI ​​model, means for requesting deletion or correction from advertisers whose fraud has been confirmed, and means for checking the advertiser's response status and imposing penalties if the advertiser does not comply. This enables rapid and accurate detection of fraudulent advertisements and content and effective countermeasure requests to advertisers. The "advertisement and content acquisition means" refers to means used to acquire new advertisements and content from web services.

[0762] An "AI model" is an artificial intelligence algorithm used to analyze data and assess its reliability.

[0763] The "reliability evaluation means" is a means for evaluating the reliability of the advertisement or content based on the acquired advertisement or content.

[0764] "Means for identifying fraudulent advertisements and content" refers to means for identifying fraudulent advertisements and content based on the analysis results of an AI model.

[0765] The "emotion data collection means" is a means for collecting emotional data generated while a user is viewing advertisements or content.

[0766] "Emotional data feedback means" refers to a means for feeding back collected emotional data to an AI model.

[0767] "Means for requesting removal or correction" are means for requesting that advertisers whose advertisements or content have been found to be fraudulent remove or correct them.

[0768] The "compliance status confirmation means" is a means for monitoring the advertiser's compliance status.

[0769] The "penalty measure" is a measure for imposing a penalty on an advertiser if the advertiser does not take the specified action.

[0770] This invention uses a "Fraudulent AI Checker" system to monitor and evaluate the reliability of advertisements and content in real time. This system aims to more accurately evaluate the fraudulent nature of advertisements and content by recognizing the emotions expressed by users while viewing or operating the content in real time and reflecting this data in the analysis of an AI model.

[0771] System Configuration

[0772] How you get ads and content

[0773] The server accesses the target web service through an API or scraping tool to obtain new advertisements and content. For example, an API is used to obtain data from a video distribution platform, and a scraping tool is used for blog sites that contain text information.

[0774] Transforming data and inputting it into AI models

[0775] The server converts the acquired advertisements and content into an analyzable format. FFMpeg is used for video data, and Google's speech recognition system is used to convert audio data into text. Once converted, the data is formatted in JSON and input into the AI ​​model.

[0776] Reliability assessment tools

[0777] AI models analyze the video, audio, and text information in ads and content to assess their credibility, comparing it with statistical benchmarks and public databases.

[0778] Emotional data collection method

[0779] The device monitors the user's reactions in real time while watching and operating the device. Using the device's camera and microphone, the device analyzes the user's facial expressions and tone of voice to collect emotional data. For example, OpenCV and DeepFace are used to obtain facial expression data from camera footage, and the Google voice analysis system is used to obtain voice data.

[0780] Emotional Data Feedback Method

[0781] The server feeds back the user's emotional data obtained from the emotion engine to the AI ​​model, which then takes the user's emotional information into account when evaluating the reliability of the AI ​​model.

[0782] How to identify fraudulent ads and content

[0783] The server receives the analysis results from the AI ​​model and identifies fraudulent ads and content. Ads and content with a reliability score below a certain threshold are recorded as fraudulent in a dedicated database.

[0784] How to request deletion or correction

[0785] The server sends a notification to the advertiser that the fraud was detected, requesting them to remove or correct the fraud. For example, an email notification is sent using SendGrid or Mailchimp.

[0786] How to check the status of response

[0787] The server monitors the advertiser's response status and checks whether the response is made within the specified deadline. The advertiser's response status is tracked using a log monitoring system such as ElasticSearch or Kibana.

[0788] Penalty measures

[0789] The server imposes penalties if the advertiser does not take the action specified by the advertiser. It has the function to automatically delete advertisements and notify users of penalties.

[0790] Examples of concrete examples and prompts

[0791] Below are some examples and prompts:

[0792] Embodiment of video distribution service

[0793] example:

[0794] 1. The server retrieves new video ads on the distribution platform via API.

[0795] 2. The server converts the video and audio data it receives into an analyzable format for input into the AI ​​model.

[0796] 3. The user's device collects emotional data while watching the ad. The emotion engine uses facial recognition and voice analysis to analyze the user's emotional state in real time.

[0797] 4. The AI ​​model evaluates the authenticity of the video ad based on the data in the ad and the emotional data sent from the device, and if the user shows an abnormal emotional response, it determines that the ad is fraudulent.

[0798] 5. The server determines the ad is fraudulent and sends a notification to the advertiser requesting its removal.

[0799] 6. After the notification, the server will check whether the advertisement is removed within the specified time limit and monitor the compliance status. If the compliance is not met, the advertisement will be automatically removed and a fine will be imposed.

[0800] Prompt Sentence Examples

[0801] "Explain how the server retrieves new video ads from a web service and how it feeds them into the AI ​​model. Use the YouTube API as a concrete example."

[0802] "Please explain the mechanism by which the device collects emotional data from users while they are watching video ads. Please specify the sensors and analysis methods used."

[0803] Please explain in detail how your AI model uses sentiment data to evaluate the trustworthiness of ads, including the specific evaluation criteria and algorithms.

[0804] In this way, the "fraudulent AI checker" system based on this invention can achieve highly accurate monitoring and rapid response, protecting end users and web service operators.

[0805] The flow of the specific processing in the second embodiment will be described with reference to FIG. 13. Detailed Description of Processing Steps

[0806] Step 1:

[0807] The server accesses the target web service (e.g., a video streaming platform) and uses an API or scraping tool to obtain new advertising and content data.

[0808] Input: API access key for web service and scraping tool settings.

[0809] Specific behavior: The server sends a request to the specified endpoint (e.g., "https: / / example.api.com / getNewAds") and downloads the metadata and content data of the new ads.

[0810] Output: The acquired advertising and content data (e.g., video files, audio files, text data) is stored in temporary storage.

[0811] Step 2:

[0812] The server converts the acquired advertisements and content into an analyzable format and inputs it into the AI ​​model.

[0813] Input: The original data of the ads and content stored.

[0814] What it does: The server splits the video file into frames using FFMpeg and converts the audio to text, which is then formatted into JSON.

[0815] Output: Data is generated in a parsable format (JSON) that is then fed into the AI ​​model.

[0816] Step 3:

[0817] The device collects the user's reactions in real time while watching and operating the device, and analyzes emotional data.

[0818] Input: Facial and voice data of the user while watching ads and content.

[0819] Specific operation: Using the device's camera and microphone, facial expressions are analyzed using OpenCV and DeepFace, and voice tone is analyzed using the Google voice analysis system.

[0820] Output: Data representing the user's emotional state (e.g., happiness, surprise, anger, sadness, etc.) is generated and sent to the server in real time.

[0821] Step 4:

[0822] The AI ​​model assesses credibility based on data and emotional data within the ad or content.

[0823] Input: Ad / content data and user sentiment data converted into an analyzable format.

[0824] How it works: The AI ​​model analyzes video, audio, and text information, compares it to statistical benchmarks and public databases, and integrates user sentiment data to perform pattern matching and anomaly detection to detect fraud.

[0825] Output: The AI ​​model generates a credibility score for each ad or piece of content.

[0826] Step 5:

[0827] The server receives the analysis results from the AI ​​model and identifies fraudulent ads and content.

[0828] Input: Analysis results of the AI ​​model (confidence score and fraud detection information).

[0829] How it works: The server checks the reliability score and lists ads and content that fall below a certain threshold. It also records this fraudulent data in a dedicated database.

[0830] Output: A list of abusive ads and content will be generated.

[0831] Step 6:

[0832] The server sends a notice to the advertiser where fraud has been confirmed, requesting removal or correction.

[0833] Input: A list of ads and content that have been identified as fraudulent, along with details about them.

[0834] What happens next: The server uses a specialized notification system (e.g., SendGrid, Mailchimp) to send an email to the advertiser requesting removal or correction. The notification will include the specific details of the fraud and a deadline for action.

[0835] Output: A notification is sent to the advertiser.

[0836] Step 7:

[0837] The server monitors the advertiser's response status and checks whether the response is made within the specified time limit. If the appropriate response is not made, a penalty will be imposed.

[0838] Input: Advertiser response reports and log data.

[0839] Specific actions: Using log monitoring systems such as ElasticSearch and Kibana, we track the advertiser's response status. If the response is not made within the deadline, we will automatically remove the ads and apply fines.

[0840] Output: Notification of the action status report and any penalties applied, if applicable.

[0841] Through the above processing steps, the system enables rapid and accurate detection and response to fraudulent advertisements and content.

[0842] (Application example 2)

[0843] Next, a description will be given of Application Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal."

[0844] In recent years, with the increase in online advertising and content, the reliability of fraudulent advertising and content has become a problem. This problem increases the risk of users receiving misinformation, leading to a decline in trust in advertising and content providers. In such situations, there is a need for systems that can detect and respond to fraud with high accuracy and speed. However, current systems have few methods for evaluating reliability using emotional data, making it difficult to detect fraud that reflects users' actual reactions.

[0845] The identification process by the identification processing unit 290 of the data processing device 12 in Application Example 2 is realized by the following means. In this invention, the server includes: means for acquiring advertisements and content; means for inputting the acquired advertisements and content into an AI model to evaluate their reliability; means having an emotion engine for collecting user emotion data and evaluating the reliability of advertisements and content based on the collected emotion data; means for identifying fraudulent advertisements and content based on the analysis results of the AI ​​model; means for requesting deletion or correction from advertisers whose fraud has been confirmed; and means for checking the advertiser's response and imposing penalties if they do not comply. This makes it possible to utilize user emotion data to more accurately evaluate the fraudulence of advertisements and content and respond quickly.

[0846] "Means for obtaining advertisements and content" refers to the technical methods for accessing and collecting advertisements and content, specifically, the means for obtaining digital information using programs such as APIs and scraping tools.

[0847] "Means for inputting acquired advertisements and content into an AI model to evaluate its reliability" refers to a series of processes for appropriately processing acquired digital information and inputting it into an AI model to analyze the reliability of that information.

[0848] "Emotion engine for collecting user emotional data" refers to technology that uses sensors and analytical algorithms to collect a user's facial expressions, tone of voice, and other data in real time to identify the user's emotional state.

[0849] "Means for evaluating the reliability of advertisements and content based on collected emotional data" refers to a mechanism that incorporates user emotional data into the analysis results and uses that data to more accurately evaluate the reliability of advertisements and content.

[0850] "Means for identifying fraudulent advertisements and content based on the analysis results of an AI model" refers to a method for automatically identifying fraudulent or suspicious advertisements and content based on data analyzed by an AI model.

[0851] "Means for requesting removal or correction from advertisers where fraud has been confirmed" refers to a method of sending a notice to an advertiser requesting the removal or correction of content when fraud is determined to have occurred.

[0852] "Measures to check the advertiser's response and impose penalties if they do not comply" refers to methods for monitoring whether the advertiser has responded to the notice and, if they do not respond, implementing penalties such as fines or automatic removal of ads.

[0853] MODE FOR CARRYING OUT THE INVENTION

[0854] An embodiment of the present invention is a system that uses user emotion data to evaluate the reliability of advertisements and content. In this system, a server mainly plays the following roles.

[0855] System Configuration

[0856] 1. How you get ads and content:

[0857] The server uses APIs or scraping tools to retrieve newly posted ads and content, giving you access to the latest digital information.

[0858] 2. Input to the AI ​​model:

[0859] The server converts the acquired advertisements and content into an analyzable format and inputs it into an AI model, which analyzes video, audio, and text information to evaluate its reliability.

[0860] 3. Emotion Engine:

[0861] The user's device is equipped with an emotion engine that uses a camera and microphone to analyze the user's facial expressions and tone of voice in real time to collect emotion data, using image processing libraries such as OpenCV and dlib.

[0862] 4. Reliability assessment measures:

[0863] The server evaluates the reliability of advertisements and content based on AI models and user emotional data. For example, if the user's emotional response is unnatural, the server determines that the advertisement or content is fraudulent.

[0864] 5. How to identify inappropriate content:

[0865] The server identifies fraudulent advertisements and content based on the analysis results of the AI ​​model, making it possible to quickly identify unreliable digital information.

[0866] 6. How to request deletion or correction:

[0867] The server sends notifications to advertisers who have been found to be fraudulent, requesting them to remove or correct the content. Notifications are efficiently handled through an automated system.

[0868] 7. How to check for compliance:

[0869] The server monitors the advertiser's response and checks whether the response is made within the specified time limit, thereby ensuring that appropriate action is taken.

[0870] 8. Means of imposing penalties:

[0871] The server will impose penalties on advertisers if they do not take appropriate action, such as automatically removing ads or applying fines.

[0872] Specific examples

[0873] Specifically, an embodiment in a video distribution service will be described.

[0874] 1. The server retrieves new video ads on the distribution platform via API.

[0875] 2. The server converts the video and audio data it receives into an analyzable format for input into the AI ​​model.

[0876] 3. The user's device collects emotional data while watching the ad. The emotion engine uses facial recognition and voice analysis to analyze the user's emotional state in real time.

[0877] 4. The AI ​​model evaluates the authenticity of the video ad based on the data in the ad and the emotional data sent from the device, and if the user shows an abnormal emotional response, it determines that the ad is fraudulent.

[0878] 5. The server determines the ad is fraudulent and sends a notification to the advertiser requesting its removal.

[0879] 6. After the notification, the server will check whether the advertisement is removed within the specified time limit and monitor the compliance status. If the compliance is not met, the advertisement will be automatically removed and a fine will be imposed.

[0880] Example prompts to input to the generative AI model

[0881] Below is an example of a prompt sentence to input to the generative AI model.

[0882] I want to create an "Emotional Ad Checker" app that evaluates the reliability of ad content. It analyzes the user's facial expressions and voice in real time using a camera and microphone, and evaluates the reliability of the ad. The process proceeds based on the following conditions:

[0883] 1. Get advertising data from the API.

[0884] 2. Capture user data with camera and microphone.

[0885] 3. Identify emotions from facial expressions and voice.

[0886] 4. Emotional data is sent to the server to evaluate the credibility of the advertisement.

[0887] The above is a specific embodiment for carrying out the present invention.

[0888] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[0889] Step 1:

[0890] The server retrieves advertisements and content. Digital information is collected from web services using APIs or scraping tools and temporarily stored. The input is data obtained via APIs or scraping tools, and the output is advertisement and content data stored in temporary storage.

[0891] Step 2:

[0892] The advertisements and content acquired by the server are converted into an analyzable format. Video, audio, and text information are arranged into the required format. The input is the data acquired in step 1, and the output is data converted into a format suitable for the AI ​​model. Specific operations include splitting video data into frames and sampling audio data.

[0893] Step 3:

[0894] The server inputs the converted ad and content data into the AI ​​model and evaluates its reliability. The AI ​​model analyzes the data and generates a reliability score. The input is the data prepared in step 2, and the output is the reliability score and analysis results. Specifically, image recognition and text analysis are performed using a neural network.

[0895] Step 4:

[0896] The device uses an emotion engine to collect real-time facial and vocal data of users watching advertisements and content. The input is video and audio data captured by a camera and microphone, and the output is analyzed emotional data. Specific operations include facial recognition and voice tone analysis.

[0897] Step 5:

[0898] The device sends the collected emotion data to the server. The input is the analyzed emotion data, and the output is the data transmission to the server. The specific operation includes executing an HTTP request to send the emotion data to the server.

[0899] Step 6:

[0900] The server combines the analysis results of the AI ​​model with the emotional data to evaluate the trustworthiness of the advertisement or content. The input is the trustworthiness score from step 3 and the emotional data from step 5, and the output is the final trustworthiness evaluation result. Specific operations include an algorithm that integrates the emotional data and the trustworthiness score and reassess the fraud risk.

[0901] Step 7:

[0902] The server identifies fraudulent ads and content based on the results of the reliability evaluation. The input is the reliability evaluation result from step 6, and the output is a list of ads and content that are determined to be fraudulent. Specific operations include a process of comparing the evaluation result with a threshold and listing those that meet the criteria for determining fraud.

[0903] Step 8:

[0904] The server sends notifications to advertisers that have been found to be fraudulent, requesting their removal or correction. The input is a list of ads or content that have been found to be fraudulent, and the output is the notifications sent to the advertisers. Specific operations include sending notifications using an automated email system.

[0905] Step 9:

[0906] The server checks the advertiser's response status and monitors whether the response is made within the specified time limit. The input is the advertiser's response status data, and the output is a log file of the response status. Specific operations include tracking the advertiser's response using a real-time monitoring system.

[0907] Step 10:

[0908] The server imposes penalties when advertisers do not take appropriate action. The input is log data on the response status, and the output is the penalty action taken. Specific actions include automatic removal of ads and application of fines.

[0909] The specific processing unit 290 transmits the result of the specific processing to the smart glasses 214. In the smart glasses 214, the control unit 46A causes the speaker 240 to output the result of the specific processing. The microphone 238 acquires audio indicating a user input regarding the result of the specific processing. The control unit 46A transmits audio data indicating the user input acquired by the microphone 238 to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the audio data.

[0910] The data generation model 58 is a so-called generative AI (Artificial Intelligence). An example of the data generation model 58 is ChatGPT (Internet Search<URL: https: / / openai.com / blog / chatgpt> ), Gemini (Internet search <url: https: gemini.google.com ?hl="ja">) and other generation AIs. The data generation model 58 is obtained by performing deep learning on a neural network. A prompt including an instruction is input to the data generation model 58, and inference data such as voice data indicating voice, text data indicating text, and image data indicating an image is also input. The data generation model 58 performs inference on the input inference data in accordance with the instruction indicated by the prompt, and outputs the inference result in a data format such as voice data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[0911] In the above embodiment, an example in which the specific processing is performed by the data processing device 12 has been given, but the technology of the present disclosure is not limited to this, and the specific processing may be performed by the smart glasses 214.

[0912] [Third embodiment]

[0913] FIG. 5 shows an example of the configuration of a data processing system 310 according to the third embodiment.

[0914] 5, the data processing system 310 includes the data processing device 12 and a headset terminal 314. An example of the data processing device 12 is a server.

[0915] The data processing device 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 is an example of a "computer" according to the technology of the present disclosure. The computer 22 includes a processor 28, a RAM 30, and a storage 32. The processor 28, the RAM 30, and the storage 32 are connected to a bus 34. The database 24 and the communication I / F 26 are also connected to the bus 34. The communication I / F 26 is connected to a network 54. Examples of the network 54 include a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[0916] The headset type terminal 314 includes a computer 36, a microphone 238, a speaker 240, a camera 42, a communication I / F 44, and a display 343. The computer 36 includes a processor 46, a RAM 48, and a storage 50. The processor 46, the RAM 48, and the storage 50 are connected to a bus 52. The microphone 238, the speaker 240, the camera 42, and the display 343 are also connected to the bus 52.

[0917] The microphone 238 receives instructions and the like from the user 20 by receiving voice uttered by the user 20. The microphone 238 captures the voice uttered by the user 20, converts the captured voice into audio data, and outputs it to the processor 46. The speaker 240 outputs audio in accordance with instructions from the processor 46.

[0918] Camera 42 is a small digital camera equipped with an optical system including a lens, aperture, and shutter, and an imaging element such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, and captures images of the surroundings of user 20 (for example, an imaging range defined by an angle of view equivalent to the field of vision of a typical healthy person).

[0919] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 are responsible for the exchange of various information between the processor 46 and the processor 28 via the network 54. The exchange of various information between the processor 46 and the processor 28 using the communication I / Fs 44 and 26 is carried out in a secure state.

[0920] Fig. 6 shows an example of the main functions of the data processing device 12 and the headset type terminal 314. As shown in Fig. 6, in the data processing device 12, a specific process is performed by the processor 28. A specific process program 56 is stored in the storage 32.

[0921] The specific processing program 56 is an example of a "program" according to the technology of the present disclosure. The processor 28 reads the specific processing program 56 from the storage 32 and executes the read specific processing program 56 on the RAM 30. The specific processing is realized by the processor 28 operating as a specific processing unit 290 in accordance with the specific processing program 56 executed on the RAM 30.

[0922] The storage 32 stores a data generation model 58 and an emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[0923] In the headset type terminal 314, a reception output process is performed by the processor 46. A reception output program 60 is stored in the storage 50. The processor 46 reads the reception output program 60 from the storage 50 and executes the read reception output program 60 on the RAM 48. The reception output process is realized by the processor 46 operating as the control unit 46A in accordance with the reception output program 60 executed on the RAM 48.

[0924] Next, a description will be given of the identification process performed by the identification processing unit 290 of the data processing device 12. In the following description, the data processing device 12 will be referred to as the "server" and the headset type terminal 314 will be referred to as the "terminal."

[0925] ---

[0926] The "Fraudulent AI Checker" system of the present invention effectively monitors advertisements and content on web services and detects false information and deep fakes generated by the use of malicious AI technology. This system is implemented primarily in a server-based configuration.

[0927] The components of this system include the following means:

[0928] 1. How you get ads and content:

[0929] The server retrieves new ads and content from the target web service using an API or scraping tool.

[0930] 2. Input to the AI ​​model:

[0931] The server converts the acquired advertisements and content into the required format and inputs them into the AI ​​model.

[0932] 3. Reliability assessment measures:

[0933] AI models analyze data within ads and content to assess their trustworthiness, covering visual, audio, and textual elements.

[0934] 4. How to identify inappropriate content:

[0935] The server identifies fraudulent advertisements and content based on the analysis results of the AI ​​model.

[0936] 5. How to request deletion or correction:

[0937] The server provides a means for sending a notice to an advertiser where fraud has been confirmed, requesting deletion or correction.

[0938] 6. How to check compatibility:

[0939] The server monitors the response status of the advertiser and checks whether the response has been made within a specified time limit.

[0940] 7. Means of imposing penalties:

[0941] To provide a means for the server to impose penalties on advertisers if they do not take appropriate action.

[0942] ---

[0943] Program processing explanation

[0944] The server first retrieves newly posted ads and content from the target web service, specifically by downloading the data using an API or scraping tool and storing it in temporary storage.

[0945] The server then converts the stored ads and content into a more easily digestible format and feeds it into an AI model that analyzes the video, audio, and text information and compares it with public sources to assess its trustworthiness.

[0946] After the AI ​​model evaluates the reliability, the server receives the analysis results and identifies fraudulent ads and content. If the reliability score falls below a certain threshold, the content is deemed fraudulent and recorded in a dedicated database.

[0947] The server will send a notice to the advertiser requesting that the advertiser remove or correct any inappropriate content that has been recorded. If the advertiser does not respond within a specified timeframe, the server will impose penalties on the advertiser, such as automatically removing the advertisement or applying a fine.

[0948] ---

[0949] Specific examples

[0950] As an example, an embodiment in a video distribution service will be described.

[0951] 1. The server calls the API to retrieve new video ads on the distribution platform.

[0952] 2. The server converts the video and audio data it receives into an analyzable format for input into the AI ​​model.

[0953] 3. An AI model analyzes data in the video ad and detects that the person in the ad was created using deepfake technology.

[0954] 4. The server determines the ad is fraudulent and sends a notification to the advertiser requesting its removal.

[0955] 5. After the server notifies the user, it checks whether the advertisement has been deleted within the specified time limit, and if the user does not comply, it automatically deletes the advertisement and imposes a penalty.

[0956] Another example is its application in blog advertising.

[0957] 1. The server retrieves newly posted ads from the blog site.

[0958] 2. The server converts the ad into text data to be input into the AI ​​model.

[0959] 3. The AI ​​model analyzes all text in the ad to detect false product information and exaggerated claims of effectiveness.

[0960] 4. The server determines that the ad is a false advertisement and sends a notice to the advertiser requesting its removal.

[0961] 5. If the advertiser does not take appropriate action after the server notifies them, the advertisements will be automatically removed from the blog site and a fine will be imposed on the advertiser.

[0962] In this way, the "Malicious AI Checker" system based on this invention provides a concrete means for quickly detecting threats posed by malicious AI use and protecting end users and web service operators.

[0963] The processing flow will be explained below.

[0964] Step 1:

[0965] The server retrieves new ads and content from the target web service. Specifically, the server uses an API or a scraping tool to download data from the web service's endpoint.

[0966] Step 2:

[0967] The server stores the acquired ads and content in temporary storage, which is fast and reliable because it is a processing stage before the data is input into the AI ​​model.

[0968] Step 3:

[0969] The server converts the stored ads and content into a more easily processed format, for example extracting video frames, transcribing audio data, or removing unnecessary tags and formatting from text data.

[0970] Step 4:

[0971] The server inputs the converted data into the AI ​​model, which then makes an API call to the AI ​​model to send the data and begin analysis.

[0972] Step 5:

[0973] AI models analyze information within ads and content, such as facial recognition and movement analysis for video, speech recognition and sentiment analysis for audio, and natural language processing for text.

[0974] Step 6:

[0975] The AI ​​model sends the analysis results to a server, which include a reliability score, suspected fraud, and specific details.

[0976] Step 7:

[0977] The server receives the analysis results from the AI ​​model and identifies fraudulent ads and content. Specifically, if the reliability score falls below a set threshold, the ad or content is deemed fraudulent.

[0978] Step 8:

[0979] The server records any advertisements or content that it determines to be fraudulent in a dedicated database, which allows it to maintain a history and be used for subsequent processing.

[0980] Step 9:

[0981] The server will then send a notice to the advertiser requesting removal or correction of the identified fraudulent content, including the reason for the fraudulent content and instructions on how to correct or remove the content.

[0982] Step 10:

[0983] The server monitors the advertiser's response status, confirms whether the advertiser has responded within the specified time limit, and records the response history.

[0984] Step 11:

[0985] If the server does not comply with the advertiser's request, it will impose penalties, such as automatic deletion of advertisements or imposition of fines.

[0986] Step 12:

[0987] The server stores all processing results and response history as logs and periodically provides reports to the web service operator, including the number of fraudulent content cases detected and the advertiser's response status.

[0988] In this way, by performing specific actions sequentially at each step, the "Fraudulent AI Checker" system effectively detects fraudulent advertisements and content and takes measures against them.

[0989] Example 1

[0990] Next, a description will be given of Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the headset type terminal 314 will be referred to as a "terminal."

[0991] Currently, advertisements and content on web services sometimes contain false information or deepfakes generated using malicious AI technology. This can lead users to make decisions based on misinformation and believe in inappropriate content or advertisements. This fraudulent content is also a major problem for web service operators, as it undermines their credibility. A system that can quickly and effectively detect and respond to these issues is needed.

[0992] The specific processing by the specific processing unit 290 of the data processing device 12 in the first embodiment is realized by the following means.

[0993] In this invention, the server includes means for acquiring advertisements and content, means for converting the acquired advertisements and content into an appropriate data format, means for inputting the converted data into a generative AI model and evaluating its reliability, means for identifying fraudulent advertisements and content based on the analysis results of the generative AI model, means for requesting deletion or correction from advertisers whose fraud has been confirmed, and means for checking the advertiser's response status and imposing penalties if the advertiser does not comply. This makes it possible to quickly detect fraudulent advertisements and content on web services and take appropriate measures.

[0994] "Advertisements and content" refers to information and media published on web services that are intended to promote purchases or provide information to users.

[0995] "Means of acquisition" refers to APIs and scraping tools used to collect advertisements and content from web services.

[0996] "Means for converting into an appropriate data format" refers to the process for converting captured advertisements or content into an analyzable format, such as image, text, audio, or video data.

[0997] A "generative AI model" is a model that uses a pre-trained artificial intelligence algorithm to evaluate the reliability and accuracy of data based on the data provided.

[0998] "Measures to assess trustworthiness" refers to the process of using generative AI models to analyze advertising and content data to determine whether the content is accurate and whether it has been generated by malicious AI technology.

[0999] "Means for identifying fraudulent advertisements and content based on analysis results" refers to the process of using the reliability assessment results output from the generative AI model to identify whether advertisements and content are fraudulent.

[1000] "Means for requesting removal or correction" refers to a notification system that requests advertisers and content providers to remove or correct advertisements or content that are deemed fraudulent.

[1001] "Measures to verify response status" refers to a process for monitoring how advertisers and content providers respond to notifications and assessing whether appropriate action has been taken within the specified timeframe.

[1002] "Measures for imposing penalties" refers to a system that implements sanctions, such as automatic removal of advertisements or imposition of fines, if advertisers or content providers do not take action within a specified deadline.

[1003] MODE FOR CARRYING OUT THE INVENTION

[1004] The "Fraudulent AI Checker" system of this invention monitors advertisements and content on web services to detect false information and deep fakes generated using malicious AI technology. This system is implemented primarily in a server-based configuration.

[1005] System Configuration

[1006] The server is the main processor and includes the following elements:

[1007] 1. How you get ads and content:

[1008] The server retrieves data from the target web service using an API or a scraping tool. An example of an API is a REST API. Examples of scraping tools include Beautiful Soup and Scrapy.

[1009] 2. Data format conversion method:

[1010] The acquired advertisements and content are converted into the appropriate data format. For example, video files are extracted as images for each frame, and audio data is converted into text. OpenCV is used for image processing, and the Google Cloud Speech-to-Text API is used for speech-to-text conversion.

[1011] 3. Input to the AI ​​model:

[1012] The server then inputs the converted data into a generative AI model, which is built using TensorFlow or PyTorch, and the model is prompted with the prompt, "Please rate the reliability of this data."

[1013] 4. Reliability assessment measures:

[1014] The generative AI model analyzes the data and assesses its reliability, specifically assessing whether the video data was generated using deepfake technology and whether the text data contains exaggeration or false information.

[1015] 5. How to identify inappropriate content:

[1016] The server receives the analysis results from the AI ​​model and identifies fraudulent ads and content based on the reliability score. If fraud is determined, the results are recorded in a dedicated database.

[1017] 6. How to request deletion or correction:

[1018] The server sends notifications to advertisers who have been found to be fraudulent, requesting them to remove or correct the content. Notifications are sent via email systems or web service notification engines.

[1019] 7. How to check for compliance:

[1020] The server monitors the advertiser's response and checks whether the appropriate action was taken within the specified time period by calling the API again and checking the data.

[1021] 8. Means of imposing penalties:

[1022] If the advertiser does not respond within the specified time frame, the server will automatically remove the ads and content and impose fines or access restrictions as necessary.

[1023] Specific examples

[1024] Examples of applications for video streaming services include:

[1025] 1. The server retrieves a new video list using the video streaming service's API.

[1026] 2. To convert the video captured by the server into an analyzable format, the video file is cut out frame by frame and the audio data is converted into text.

[1027] 3. The server inputs the converted data into the deepfake detection AI model, entering the prompt "Please rate the reliability of this data."

[1028] 4. The AI ​​model analyzes the video frames and detects that they are deepfakes.

[1029] 5. The server determines that the video is fraudulent and records it in the database.

[1030] 6. The server sends a removal request notification to the advertiser.

[1031] 7. The server checks whether the video has been deleted within the specified time limit.

[1032] 8. If the server does not respond, the video will be automatically deleted and penalties will be applied.

[1033] In this way, the "Fraudulent AI Checker" system based on this invention provides a concrete means to quickly detect the threat of false information and deep fakes caused by fraudulent AI technology and protect users and web service operators.

[1034] The flow of the identification process in the first embodiment will be described with reference to FIG.

[1035] Step 1:

[1036] Acquiring advertisements and content

[1037] The server retrieves ads and content from the target web service. Specifically, it collects data using an API or scraping tool. The retrieved data is then stored in temporary storage. The input is the web service's API endpoint or scraping tool settings, and the output is the retrieved ad and content data. This ensures that the latest ads and content are available within the system.

[1038] Step 2:

[1039] Data format conversion

[1040] The server converts the advertisements and content it acquires into an appropriate data format. Specifically, for video files, images are extracted frame by frame, and audio data is converted into text. The text information is used as is. The input is the acquired raw data, and the output is data converted into a format that can be analyzed by the generative AI model. This prepares data suitable for the AI ​​model.

[1041] Step 3:

[1042] Input to the AI ​​model

[1043] The server inputs the converted data into a generative AI model, which is built using TensorFlow or PyTorch. The model receives a prompt, "Please rate the reliability of this data." The input is the converted data and the prompt, and the output is the result of the AI ​​model's reliability assessment. This starts the process of assessing the reliability of the data.

[1044] Step 4:

[1045] Reliability assessment

[1046] The generative AI model analyzes the provided data and assesses its reliability. Specifically, it assesses whether deepfake technology has been used in video data and whether false or exaggerated information is included in text data. The input is a set of data to the AI ​​model, and the output is a reliability score or analysis results. This prepares the model for the next step based on the analysis results.

[1047] Step 5:

[1048] Identifying fraudulent content

[1049] The server receives the analysis results of the AI ​​model and identifies fraudulent ads and content based on the reliability score. A threshold score is set for what is deemed fraudulent, and content below this threshold is identified as fraudulent. The input is the reliability score and analysis results, and the output is a list of fraudulent content. This allows specific fraudulent content to be identified and addressed.

[1050] Step 6:

[1051] Requests for removal or correction

[1052] The server sends a notification to the advertiser where the fraud has been confirmed, requesting removal or correction. The notification is sent via an email system or a web service notification engine. The input is a list of fraudulent content and the advertiser's contact information, and the output is the notification sent. This allows the appropriate instructions to be given to the advertiser.

[1053] Step 7:

[1054] Check compatibility

[1055] The server monitors the advertiser's response and verifies whether appropriate action has been taken within the specified time limit. The server then calls the API again to retrieve data and verify whether the original inappropriate content has been corrected or deleted. The input is the re-retrieved data, and the output is the confirmation result of whether action has been taken. This makes it possible to understand the advertiser's response status.

[1056] Step 8:

[1057] Imposing a penalty

[1058] If the server does not take action within the specified deadline, it imposes penalties on the advertiser or content provider. Specifically, it automatically deletes the fraudulent ads and content, and imposes fines or access restrictions as necessary. The input is the response result and penalty conditions, and the output is the executed penalty. This allows inappropriate responses to be punished quickly, maintaining the reliability of the system.

[1059] (Application example 1)

[1060] Next, a description will be given of Application Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the headset type terminal 314 will be referred to as a "terminal."

[1061] In recent years, the use of malicious AI technology to create false information and deep fakes in advertisements and content has increased, increasing the likelihood of ordinary consumers being misled or harmed. There is a need for a means to quickly and effectively detect, correct, or remove such fraudulent advertisements and content. There is also a need for a system that can immediately warn viewers of the risk of fraudulent advertisements and urge them to be careful.

[1062] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 1 is realized by the following means.

[1063] In this invention, the server includes means for acquiring advertisements and content, means for inputting the acquired advertisements and content into an AI model and evaluating their reliability, means for identifying fraudulent advertisements and content based on the analysis results of the AI ​​model, means for requesting deletion or correction from advertisers whose fraud has been confirmed, means for checking the advertiser's response status and imposing penalties if the advertiser does not comply, means for automatically capturing advertisements viewed by users, means for analyzing the captured advertisements and evaluating their reliability, means for notifying users of a warning when a fraudulent advertisement is detected, and means for sending a warning to the advertiser as well. This improves the reliability of advertisements viewed by users and enables rapid detection and response to fraudulent advertisements.

[1064] "Means of obtaining advertisements and content" refers to the function of obtaining advertisements and content on web services using APIs or scraping tools.

[1065] "Means of inputting into an AI model and evaluating reliability" refers to a function that converts acquired advertisements and content into an appropriate format and inputs it into an AI model to evaluate its reliability.

[1066] "Means for identifying fraudulent advertisements and content" refers to a function that determines the reliability of advertisements and content based on the analysis results of an AI model and identifies those that contain fraudulent elements.

[1067] "Means to request removal or correction" is a function that sends a notification to the advertiser requesting removal or correction if fraudulent advertisements or content are confirmed.

[1068] "Means to check the advertiser's response status and impose penalties if they do not comply" is a function that checks whether the advertiser has responded appropriately to requests for deletion or correction, and imposes penalties if they do not comply.

[1069] "Means for automatically capturing advertisements viewed by users" refers to a function that automatically captures screenshots and URLs of advertisements viewed by users.

[1070] "Means for analyzing captured advertisements and evaluating their reliability" refers to a function that converts the captured advertisement data into an appropriate format and inputs it into an AI model to evaluate its reliability.

[1071] "Means for notifying users of a warning when fraudulent advertising is detected" is a function that displays a warning to users when fraudulent advertising is detected.

[1072] The "means for sending a warning notice to the advertiser" is a function that, when a fraudulent advertisement is detected, sends a warning notice to the advertiser that provided the advertisement.

[1073] The "Fraudulent AI Checker" system of this invention evaluates the reliability of advertisements and content viewed by users and detects false information and deep fakes. This system is implemented in a server-based configuration.

[1074] Components

[1075] 1. Server

[1076] 2. User device (smartphone)

[1077] 3. AI Model

[1078] Program structure and operation

[1079] 1. Acquiring advertisements and content

[1080] The server periodically retrieves new advertisements and content from the web service using an API or a scraping tool. The system also includes a function to automatically capture screenshots and URLs of advertisements viewed by the user's device. Smartphone screen capture APIs (such as Android's MediaProjection API and iOS's UIScreenshotService) are used for this purpose.

[1081] 2. Data conversion and input to AI models

[1082] The acquired advertisements and content are converted into an analyzable format by the server. For example, an image recognition API (such as Google Cloud Vision API) is used to convert the data into text or video data, which is then input into an AI model.

[1083] 3. Reliability evaluation

[1084] AI models (generative AI models built with TensorFlow, PyTorch, etc.) analyze ad and content data and evaluate their trustworthiness, with the aim of detecting deep fakes and false information.

[1085] 4. Identifying and warning fraudulent ads

[1086] Based on the analysis results of the AI ​​model, the server identifies fraudulent ads and content. If fraud is detected, the server displays a warning to the user. It also automatically sends a warning to the advertiser, requesting that the content be removed or corrected.

[1087] 5. Response confirmation and penalties

[1088] The server monitors whether the advertiser has responded appropriately to the deletion or correction request. If the request is not responded to within the specified time frame, the server will automatically delete the ad and impose a penalty on the advertiser. This notification is sent using services such as Firebase Cloud Messaging and SendGrid.

[1089] Specific examples

[1090] Application to video advertising

[1091] The server retrieves new video ads using the video distribution platform's API.

[1092] The acquired video advertisements are converted into a format that can be analyzed by the AI ​​model (video, audio).

[1093] AI model detects deepfakes in video ads.

[1094] The server determined the content to be fraudulent and sent a notification to the advertiser requesting its removal.

[1095] A warning message appears on the user's device stating, "This ad may be false."

[1096] If the advertiser does not comply, the server will automatically remove the video and impose a penalty.

[1097] Application to blog advertising

[1098] The server retrieves new ads from the blog site using a scraping tool.

[1099] Convert it into text data and input it into the AI ​​model.

[1100] AI models detect false product information and exaggerated claims.

[1101] The server determines the ad is fraudulent and notifies the advertiser to remove it.

[1102] A warning message appears on the user's device stating, "This ad may be false."

[1103] If advertisers do not comply, their ads will be automatically removed from blog sites and fines will be imposed.

[1104] Prompt Sentence Examples

[1105] "Enter your ad text and we'll check it for false or exaggerated information."

[1106] As described above, the system provides a concrete means to quickly assess the trustworthiness of advertisements and content, and protect users from false information and fraudulent advertisements.

[1107] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[1108] Step 1: Getting ads and content

[1109] The server periodically obtains new advertisements and content from a web service using an API or scraping tool. The input is the web service's URL or API endpoint, and the output is the newly obtained advertisement or content data. Specifically, the server calls the API and receives JSON-formatted advertisement data as a response. The server also uses a scraping tool to extract advertisement information from HTML pages and temporarily store it.

[1110] Step 2: Capture the ads users see

[1111] The device automatically captures a screenshot or URL of the ad the user is viewing. The input is the ad screen being viewed, and the output is the captured screenshot or URL. Specifically, the device's screen capture API (for example, Android's MediaProjection API or iOS's UIScreenshotService) is used to capture an image of the ad being viewed.

[1112] Step 3: Data conversion

[1113] The server converts the acquired ad and content data into an analyzable format. The input is screenshots and ad data, and the output is text or video data. Specifically, it uses an image recognition API (such as Google Cloud Vision API) to extract text from ad screenshots and convert it into the required format. In the case of video data, it uses a video analysis algorithm to split it into analyzable frames.

[1114] Step 4: Input to the AI ​​model

[1115] The server inputs the converted data into an AI model. The input is text data or video data, and the output is a reliability evaluation result. Specifically, the data is input into a generative AI model built with TensorFlow or PyTorch, and reliability evaluation is performed. The AI ​​model performs analysis based on a pre-trained dataset.

[1116] Step 5: Reliability assessment

[1117] The server receives the analysis results of the AI ​​model and evaluates the reliability of the advertisements and content. The input is the reliability score obtained from the AI ​​model, and the output is the reliability evaluation result of the advertisement or content. Specifically, if the reliability score falls below a certain threshold, it is determined to be fraudulent and recorded in a dedicated database.

[1118] Step 6: Identifying and warning fraudulent ads

[1119] The server displays a warning to the user about ads or content that is determined to be fraudulent. The input is the reliability assessment result, and the output is a warning to the user. Specifically, it uses Firebase Cloud Messaging and the device's notification function to display a message on the user's device saying, "This ad may be false."

[1120] Step 7: Notify advertisers

[1121] The server automatically sends a warning to advertisers who have been found to be fraudulent, requesting them to remove or correct the content. The input is the reliability assessment result and the advertiser's contact information, and the output is the notification sent. Specifically, an email service such as SendGrid is used to send an email to the advertiser requesting removal.

[1122] Step 8: Action Verification and Penalties

[1123] The server monitors whether the advertiser has responded appropriately to requests for deletion or correction. The input is the advertiser's response status, and the output is the monitoring results and penalty processing. Specifically, if no response is made, the server will automatically delete the advertisement and take steps to impose a penalty on the advertiser.

[1124] This process strengthens users' defenses against untrustworthy ads and encourages advertisers to take action quickly.

[1125] Furthermore, an emotion engine that estimates the user's emotion may be further combined. That is, the identification processing unit 290 may estimate the user's emotion using the emotion identification model 59, and perform identification processing using the user's emotion.

[1126] ---

[1127] By combining an emotion engine with the "Fraudulent AI Checker" system of the present invention, it is possible to monitor user reactions and further improve the reliability of advertisements and content based on the results. This system recognizes the emotions expressed by users while viewing or operating the content in real time, and reflects this data in the analysis of the AI ​​model, thereby more accurately evaluating the fraudulent nature of advertisements and content.

[1128] The components of this system include the following means:

[1129] 1. How you get ads and content:

[1130] The server retrieves new ads and content from the target web service using an API or scraping tool.

[1131] 2. Input to the AI ​​model:

[1132] The server converts the acquired advertisements and content into the required format and inputs them into the AI ​​model.

[1133] 3. Reliability assessment measures:

[1134] AI models analyze data within ads and content to assess their trustworthiness, including visual, audio, and textual elements.

[1135] 4. How to identify inappropriate content:

[1136] The server identifies fraudulent advertisements and content based on the analysis results of the AI ​​model.

[1137] 5. How to request deletion or correction:

[1138] The server provides a means for sending a notice to an advertiser where fraud has been confirmed, requesting deletion or correction.

[1139] 6. How to check compatibility:

[1140] The server monitors the response status of the advertiser and checks whether the response has been made within a specified time limit.

[1141] 7. Means of imposing penalties:

[1142] To provide a means for the server to impose penalties on advertisers if they do not take appropriate action.

[1143] 8. Emotion Engine:

[1144] The device monitors the user's reactions to what they are watching and doing in real time, collecting data. The emotion engine uses sensors such as cameras and microphones to analyze the user's facial expressions and tone of voice.

[1145] 9. Emotional data feedback methods:

[1146] The server feeds back the user's emotional data obtained from the emotion engine to the AI ​​model, allowing the user's actual reaction to be taken into account when evaluating trustworthiness.

[1147] ---

[1148] Program processing explanation

[1149] The server first retrieves newly posted ads and content from the target web service, specifically by downloading the data using an API or scraping tool and storing it in temporary storage.

[1150] The server then converts the stored ads and content into a more easily digestible format and feeds it into an AI model that analyzes the video, audio, and text information and compares it with public sources to assess its trustworthiness.

[1151] The server receives emotional data from the emotion engine on each device. When a user watches an advertisement or watches content and shows some kind of reaction to it, the emotion engine analyzes their facial expressions and tone of voice to determine their emotional state.

[1152] The AI ​​model takes emotional data into account when assessing the trustworthiness of ads and content. In addition to regular data analysis, the AI ​​model checks whether the user's emotional data indicates unnatural reactions, improving the accuracy of the assessment.

[1153] The server receives the analysis results from the AI ​​model and identifies fraudulent ads and content. If the reliability score falls below a certain threshold, the content is deemed fraudulent and recorded in a dedicated database.

[1154] The server will send a notice to the advertiser requesting that the advertiser remove or correct any inappropriate content that has been recorded. If the advertiser does not respond within a specified timeframe, the server will impose penalties on the advertiser, such as automatically removing the advertisement or applying a fine.

[1155] ---

[1156] Specific examples

[1157] As an example, an embodiment in a video distribution service will be described.

[1158] 1. The server retrieves new video ads on the distribution platform via API.

[1159] 2. The server converts the video and audio data it receives into an analyzable format for input into the AI ​​model.

[1160] 3. The user's device collects emotional data while watching the ad. The emotion engine uses facial recognition and voice analysis to analyze the user's emotional state in real time.

[1161] 4. The AI ​​model evaluates the authenticity of the video ad based on the data in the ad and the emotional data sent from the device, and if the user shows an abnormal emotional response, it determines that the ad is fraudulent.

[1162] 5. The server determines the ad is fraudulent and sends a notification to the advertiser requesting its removal.

[1163] 6. After the notification, the server will check whether the advertisement is removed within the specified time limit and monitor the compliance status. If the compliance is not met, the advertisement will be automatically removed and a fine will be imposed.

[1164] Another example is its application in blog advertising.

[1165] 1. The server retrieves newly posted ads from the blog site.

[1166] 2. The server converts the ad into text data to be input into the AI ​​model.

[1167] 3. The user's device collects emotional data while viewing the ad. The emotion engine analyzes the user's facial expressions and tone of voice to identify their emotional state.

[1168] 4. The AI ​​model analyzes all the text in the ad, takes into account emotional data, and if it finds a high likelihood of causing misunderstanding or distrust in the user, it will determine that the ad is false.

[1169] 5. The server determines that the ad is a false advertisement and sends a notice to the advertiser requesting its removal.

[1170] 6. If the advertiser does not take appropriate action after the server notifies them, the advertisements will be automatically removed from the blog site and a fine will be imposed on the advertiser.

[1171] In this way, the "Fraudulent AI Checker" system based on this invention utilizes an emotion engine to quickly and accurately detect threats posed by fraudulent AI use, providing a concrete means for protecting end users and web service operators.

[1172] The processing flow will be explained below.

[1173] Step 1:

[1174] The server obtains new ads and content from the target web service. Specifically, the server downloads the latest ad and content data from the web service endpoint using an API or scraping tool.

[1175] Step 2:

[1176] The server stores the retrieved advertisements and content in temporary storage, allowing the data to proceed to the next processing step without loss.

[1177] Step 3:

[1178] The server converts the stored ads and content into a more easily processed format, for example extracting video frames, transcribing audio data, or removing unnecessary tags and formatting from text data.

[1179] Step 4:

[1180] The server inputs the converted data into the AI ​​model, which then makes an API call to the AI ​​model to send the analyzed data, starting processing.

[1181] Step 5:

[1182] When a user's device starts displaying an advertisement or content, the device's emotion engine analyzes the user's facial expressions and voice in real time, for example, capturing the user's facial expressions with a camera and analyzing the tone of the voice with a microphone.

[1183] Step 6:

[1184] The device then sends the analyzed emotional data to the server, which includes the user's reactions to the displayed advertisements and content (such as surprise, disbelief, or anger).

[1185] Step 7:

[1186] AI models analyze data within ads and content to assess their trustworthiness, taking into account video, audio, and text information, as well as emotional data transmitted from the device.

[1187] Step 8:

[1188] The server receives the analysis results from the AI ​​model and identifies fraudulent ads and content. If the reliability score falls below a set threshold, the ad or content is deemed fraudulent.

[1189] Step 9:

[1190] The server records any ads or content that it determines to be fraudulent in a dedicated database, allowing it to maintain a history and detailed information about the fraudulent content.

[1191] Step 10:

[1192] The server will then send a notice to the advertiser requesting removal or correction of the fraudulent content, including the reason for the fraudulent content and how to correct it.

[1193] Step 11:

[1194] The server monitors the advertiser's response status, checks whether the advertiser has responded appropriately within the specified time period, and records the response details and their history.

[1195] Step 12:

[1196] The server will penalize advertisers if they fail to comply, automatically removing any abusive ads or content and notifying them of penalties if necessary.

[1197] Step 13:

[1198] The server stores all processing results and response history as logs and periodically provides reports to the web service operator, including the number of fraudulent content detected and the advertiser's response status.

[1199] ---

[1200] With this processing flow, the "Fraudulent AI Checker" system based on this invention utilizes an emotion engine to effectively identify fraudulent advertisements and content, providing a concrete means for protecting users and web service operators.

[1201] Example 2

[1202] Next, a description will be given of Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the headset type terminal 314 will be referred to as a "terminal."

[1203] The distribution of fraudulent information in modern advertising and content has become a problem. In particular, advertisements and content containing false information are rampant on the Internet, resulting in confusion for users and a loss of trust in web services. Therefore, it is necessary to quickly and accurately detect such fraudulent advertisements and content and take appropriate measures.

[1204] The identification process by the identification processing unit 290 of the data processing device 12 in Example 2 is realized by the following means. In this invention, the server includes means for acquiring advertisements and content, means for inputting the acquired advertisements and content into an AI model and evaluating their reliability, means for identifying fraudulent advertisements and content based on the analysis results of the AI ​​model, means for collecting emotional data of users viewing content on the user's terminal and feeding the data back to the AI ​​model, means for requesting deletion or correction from advertisers whose fraud has been confirmed, and means for checking the advertiser's response status and imposing penalties if the advertiser does not comply. This enables rapid and accurate detection of fraudulent advertisements and content and effective countermeasure requests to advertisers. The "advertisement and content acquisition means" refers to means used to acquire new advertisements and content from web services.

[1205] An "AI model" is an artificial intelligence algorithm used to analyze data and assess its reliability.

[1206] The "reliability evaluation means" is a means for evaluating the reliability of the advertisement or content based on the acquired advertisement or content.

[1207] "Means for identifying fraudulent advertisements and content" refers to means for identifying fraudulent advertisements and content based on the analysis results of an AI model.

[1208] The "emotion data collection means" is a means for collecting emotional data generated while a user is viewing advertisements or content.

[1209] "Emotional data feedback means" refers to a means for feeding back collected emotional data to an AI model.

[1210] "Means for requesting removal or correction" are means for requesting that advertisers whose advertisements or content have been found to be fraudulent remove or correct them.

[1211] The "compliance status confirmation means" is a means for monitoring the advertiser's compliance status.

[1212] The "penalty measure" is a measure for imposing a penalty on an advertiser if the advertiser does not take the specified action.

[1213] This invention uses a "Fraudulent AI Checker" system to monitor and evaluate the reliability of advertisements and content in real time. This system aims to more accurately evaluate the fraudulent nature of advertisements and content by recognizing the emotions expressed by users while viewing or operating the content in real time and reflecting this data in the analysis of an AI model.

[1214] System Configuration

[1215] How you get ads and content

[1216] The server accesses the target web service through an API or scraping tool to obtain new advertisements and content. For example, an API is used to obtain data from a video distribution platform, and a scraping tool is used for blog sites that contain text information.

[1217] Transforming data and inputting it into AI models

[1218] The server converts the acquired advertisements and content into an analyzable format. FFMpeg is used for video data, and Google's speech recognition system is used to convert audio data into text. Once converted, the data is formatted in JSON and input into the AI ​​model.

[1219] Reliability assessment tools

[1220] AI models analyze the video, audio, and text information in ads and content to assess their credibility, comparing it with statistical benchmarks and public databases.

[1221] Emotional data collection method

[1222] The device monitors the user's reactions in real time while watching and operating the device. Using the device's camera and microphone, the device analyzes the user's facial expressions and tone of voice to collect emotional data. For example, OpenCV and DeepFace are used to obtain facial expression data from camera footage, and the Google voice analysis system is used to obtain voice data.

[1223] Emotional Data Feedback Method

[1224] The server feeds back the user's emotional data obtained from the emotion engine to the AI ​​model, which then takes the user's emotional information into account when evaluating the reliability of the AI ​​model.

[1225] How to identify fraudulent ads and content

[1226] The server receives the analysis results from the AI ​​model and identifies fraudulent ads and content. Ads and content with a reliability score below a certain threshold are recorded as fraudulent in a dedicated database.

[1227] How to request deletion or correction

[1228] The server sends a notification to the advertiser that the fraud was detected, requesting them to remove or correct the fraud. For example, an email notification is sent using SendGrid or Mailchimp.

[1229] How to check the status of response

[1230] The server monitors the advertiser's response status and checks whether the response is made within the specified deadline. The advertiser's response status is tracked using a log monitoring system such as ElasticSearch or Kibana.

[1231] Penalty measures

[1232] The server imposes penalties if the advertiser does not take the action specified by the advertiser. It has the function to automatically delete advertisements and notify users of penalties.

[1233] Examples of concrete examples and prompts

[1234] Below are some examples and prompts:

[1235] Embodiment of video distribution service

[1236] example:

[1237] 1. The server retrieves new video ads on the distribution platform via API.

[1238] 2. The server converts the video and audio data it receives into an analyzable format for input into the AI ​​model.

[1239] 3. The user's device collects emotional data while watching the ad. The emotion engine uses facial recognition and voice analysis to analyze the user's emotional state in real time.

[1240] 4. The AI ​​model evaluates the authenticity of the video ad based on the data in the ad and the emotional data sent from the device, and if the user shows an abnormal emotional response, it determines that the ad is fraudulent.

[1241] 5. The server determines the ad is fraudulent and sends a notification to the advertiser requesting its removal.

[1242] 6. After the notification, the server will check whether the advertisement is removed within the specified time limit and monitor the compliance status. If the compliance is not met, the advertisement will be automatically removed and a fine will be imposed.

[1243] Prompt Sentence Examples

[1244] "Explain how the server retrieves new video ads from a web service and how it feeds them into the AI ​​model. Use the YouTube API as a concrete example."

[1245] "Please explain the mechanism by which the device collects emotional data from users while they are watching video ads. Please specify the sensors and analysis methods used."

[1246] Please explain in detail how your AI model uses sentiment data to evaluate the trustworthiness of ads, including the specific evaluation criteria and algorithms.

[1247] In this way, the "fraudulent AI checker" system based on this invention can achieve highly accurate monitoring and rapid response, protecting end users and web service operators.

[1248] The flow of the specific processing in the second embodiment will be described with reference to FIG. 13. Detailed Description of Processing Steps

[1249] Step 1:

[1250] The server accesses the target web service (e.g., a video streaming platform) and uses an API or scraping tool to obtain new advertising and content data.

[1251] Input: API access key for web service and scraping tool settings.

[1252] Specific behavior: The server sends a request to the specified endpoint (e.g., "https: / / example.api.com / getNewAds") and downloads the metadata and content data of the new ads.

[1253] Output: The acquired advertising and content data (e.g., video files, audio files, text data) is stored in temporary storage.

[1254] Step 2:

[1255] The server converts the acquired advertisements and content into an analyzable format and inputs it into the AI ​​model.

[1256] Input: The original data of the ads and content stored.

[1257] What it does: The server splits the video file into frames using FFMpeg and converts the audio to text, which is then formatted into JSON.

[1258] Output: Data is generated in a parsable format (JSON) that is then fed into the AI ​​model.

[1259] Step 3:

[1260] The device collects the user's reactions in real time while watching and operating the device, and analyzes emotional data.

[1261] Input: Facial and voice data of the user while watching ads and content.

[1262] Specific operation: Using the device's camera and microphone, facial expressions are analyzed using OpenCV and DeepFace, and voice tone is analyzed using the Google voice analysis system.

[1263] Output: Data representing the user's emotional state (e.g., happiness, surprise, anger, sadness, etc.) is generated and sent to the server in real time.

[1264] Step 4:

[1265] The AI ​​model assesses credibility based on data and emotional data within the ad or content.

[1266] Input: Ad / content data and user sentiment data converted into an analyzable format.

[1267] How it works: The AI ​​model analyzes video, audio, and text information, compares it to statistical benchmarks and public databases, and integrates user sentiment data to perform pattern matching and anomaly detection to detect fraud.

[1268] Output: The AI ​​model generates a credibility score for each ad or piece of content.

[1269] Step 5:

[1270] The server receives the analysis results from the AI ​​model and identifies fraudulent ads and content.

[1271] Input: Analysis results of the AI ​​model (confidence score and fraud detection information).

[1272] How it works: The server checks the reliability score and lists ads and content that fall below a certain threshold. It also records this fraudulent data in a dedicated database.

[1273] Output: A list of abusive ads and content will be generated.

[1274] Step 6:

[1275] The server sends a notice to the advertiser where fraud has been confirmed, requesting removal or correction.

[1276] Input: A list of ads and content that have been identified as fraudulent, along with details about them.

[1277] What happens next: The server uses a specialized notification system (e.g., SendGrid, Mailchimp) to send an email to the advertiser requesting removal or correction. The notification will include the specific details of the fraud and a deadline for action.

[1278] Output: A notification is sent to the advertiser.

[1279] Step 7:

[1280] The server monitors the advertiser's response status and checks whether the response is made within the specified time limit. If the appropriate response is not made, a penalty will be imposed.

[1281] Input: Advertiser response reports and log data.

[1282] Specific actions: Using log monitoring systems such as ElasticSearch and Kibana, we track the advertiser's response status. If the response is not made within the deadline, we will automatically remove the ads and apply fines.

[1283] Output: Notification of the action status report and any penalties applied, if applicable.

[1284] Through the above processing steps, the system enables rapid and accurate detection and response to fraudulent advertisements and content.

[1285] (Application example 2)

[1286] Next, a description will be given of Application Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the headset type terminal 314 will be referred to as a "terminal."

[1287] In recent years, with the increase in online advertising and content, the reliability of fraudulent advertising and content has become a problem. This problem increases the risk of users receiving misinformation, leading to a decline in trust in advertising and content providers. In such situations, there is a need for systems that can detect and respond to fraud with high accuracy and speed. However, current systems have few methods for evaluating reliability using emotional data, making it difficult to detect fraud that reflects users' actual reactions.

[1288] The identification process by the identification processing unit 290 of the data processing device 12 in Application Example 2 is realized by the following means. In this invention, the server includes: means for acquiring advertisements and content; means for inputting the acquired advertisements and content into an AI model to evaluate their reliability; means having an emotion engine for collecting user emotion data and evaluating the reliability of advertisements and content based on the collected emotion data; means for identifying fraudulent advertisements and content based on the analysis results of the AI ​​model; means for requesting deletion or correction from advertisers whose fraud has been confirmed; and means for checking the advertiser's response and imposing penalties if they do not comply. This makes it possible to utilize user emotion data to more accurately evaluate the fraudulence of advertisements and content and respond quickly.

[1289] "Means for obtaining advertisements and content" refers to the technical methods for accessing and collecting advertisements and content, specifically, the means for obtaining digital information using programs such as APIs and scraping tools.

[1290] "Means for inputting acquired advertisements and content into an AI model to evaluate its reliability" refers to a series of processes for appropriately processing acquired digital information and inputting it into an AI model to analyze the reliability of that information.

[1291] "Emotion engine for collecting user emotional data" refers to technology that uses sensors and analytical algorithms to collect a user's facial expressions, tone of voice, and other data in real time to identify the user's emotional state.

[1292] "Means for evaluating the reliability of advertisements and content based on collected emotional data" refers to a mechanism that incorporates user emotional data into the analysis results and uses that data to more accurately evaluate the reliability of advertisements and content.

[1293] "Means for identifying fraudulent advertisements and content based on the analysis results of an AI model" refers to a method for automatically identifying fraudulent or suspicious advertisements and content based on data analyzed by an AI model.

[1294] "Means for requesting removal or correction from advertisers where fraud has been confirmed" refers to a method of sending a notice to an advertiser requesting the removal or correction of content when fraud is determined to have occurred.

[1295] "Measures to check the advertiser's response and impose penalties if they do not comply" refers to methods for monitoring whether the advertiser has responded to the notice and, if they do not respond, implementing penalties such as fines or automatic removal of ads.

[1296] MODE FOR CARRYING OUT THE INVENTION

[1297] An embodiment of the present invention is a system that uses user emotion data to evaluate the reliability of advertisements and content. In this system, a server mainly plays the following roles.

[1298] System Configuration

[1299] 1. How you get ads and content:

[1300] The server uses APIs or scraping tools to retrieve newly posted ads and content, giving you access to the latest digital information.

[1301] 2. Input to the AI ​​model:

[1302] The server converts the acquired advertisements and content into an analyzable format and inputs it into an AI model, which analyzes video, audio, and text information to evaluate its reliability.

[1303] 3. Emotion Engine:

[1304] The user's device is equipped with an emotion engine that uses a camera and microphone to analyze the user's facial expressions and tone of voice in real time to collect emotion data, using image processing libraries such as OpenCV and dlib.

[1305] 4. Reliability assessment measures:

[1306] The server evaluates the reliability of advertisements and content based on AI models and user emotional data. For example, if the user's emotional response is unnatural, the server determines that the advertisement or content is fraudulent.

[1307] 5. How to identify inappropriate content:

[1308] The server identifies fraudulent advertisements and content based on the analysis results of the AI ​​model, making it possible to quickly identify unreliable digital information.

[1309] 6. How to request deletion or correction:

[1310] The server sends notifications to advertisers who have been found to be fraudulent, requesting them to remove or correct the content. Notifications are efficiently handled through an automated system.

[1311] 7. How to check for compliance:

[1312] The server monitors the advertiser's response and checks whether the response is made within the specified time limit, thereby ensuring that appropriate action is taken.

[1313] 8. Means of imposing penalties:

[1314] The server will impose penalties on advertisers if they do not take appropriate action, such as automatically removing ads or applying fines.

[1315] Specific examples

[1316] Specifically, an embodiment in a video distribution service will be described.

[1317] 1. The server retrieves new video ads on the distribution platform via API.

[1318] 2. The server converts the video and audio data it receives into an analyzable format for input into the AI ​​model.

[1319] 3. The user's device collects emotional data while watching the ad. The emotion engine uses facial recognition and voice analysis to analyze the user's emotional state in real time.

[1320] 4. The AI ​​model evaluates the authenticity of the video ad based on the data in the ad and the emotional data sent from the device, and if the user shows an abnormal emotional response, it determines that the ad is fraudulent.

[1321] 5. The server determines the ad is fraudulent and sends a notification to the advertiser requesting its removal.

[1322] 6. After the notification, the server will check whether the advertisement is removed within the specified time limit and monitor the compliance status. If the compliance is not met, the advertisement will be automatically removed and a fine will be imposed.

[1323] Example prompts to input to the generative AI model

[1324] Below is an example of a prompt sentence to input to the generative AI model.

[1325] I want to create an "Emotional Ad Checker" app that evaluates the reliability of ad content. It analyzes the user's facial expressions and voice in real time using a camera and microphone, and evaluates the reliability of the ad. The process proceeds based on the following conditions:

[1326] 1. Get advertising data from the API.

[1327] 2. Capture user data with camera and microphone.

[1328] 3. Identify emotions from facial expressions and voice.

[1329] 4. Emotional data is sent to the server to evaluate the credibility of the advertisement.

[1330] The above is a specific embodiment for carrying out the present invention.

[1331] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[1332] Step 1:

[1333] The server retrieves advertisements and content. Digital information is collected from web services using APIs or scraping tools and temporarily stored. The input is data obtained via APIs or scraping tools, and the output is advertisement and content data stored in temporary storage.

[1334] Step 2:

[1335] The advertisements and content acquired by the server are converted into an analyzable format. Video, audio, and text information are arranged into the required format. The input is the data acquired in step 1, and the output is data converted into a format suitable for the AI ​​model. Specific operations include splitting video data into frames and sampling audio data.

[1336] Step 3:

[1337] The server inputs the converted ad and content data into the AI ​​model and evaluates its reliability. The AI ​​model analyzes the data and generates a reliability score. The input is the data prepared in step 2, and the output is the reliability score and analysis results. Specifically, image recognition and text analysis are performed using a neural network.

[1338] Step 4:

[1339] The device uses an emotion engine to collect real-time facial and vocal data of users watching advertisements and content. The input is video and audio data captured by a camera and microphone, and the output is analyzed emotional data. Specific operations include facial recognition and voice tone analysis.

[1340] Step 5:

[1341] The device sends the collected emotion data to the server. The input is the analyzed emotion data, and the output is the data transmission to the server. The specific operation includes executing an HTTP request to send the emotion data to the server.

[1342] Step 6:

[1343] The server combines the analysis results of the AI ​​model with the emotional data to evaluate the trustworthiness of the advertisement or content. The input is the trustworthiness score from step 3 and the emotional data from step 5, and the output is the final trustworthiness evaluation result. Specific operations include an algorithm that integrates the emotional data and the trustworthiness score and reassess the fraud risk.

[1344] Step 7:

[1345] The server identifies fraudulent ads and content based on the results of the reliability evaluation. The input is the reliability evaluation result from step 6, and the output is a list of ads and content that are determined to be fraudulent. Specific operations include a process of comparing the evaluation result with a threshold and listing those that meet the criteria for determining fraud.

[1346] Step 8:

[1347] The server sends notifications to advertisers that have been found to be fraudulent, requesting their removal or correction. The input is a list of ads or content that have been found to be fraudulent, and the output is the notifications sent to the advertisers. Specific operations include sending notifications using an automated email system.

[1348] Step 9:

[1349] The server checks the advertiser's response status and monitors whether the response is made within the specified time limit. The input is the advertiser's response status data, and the output is a log file of the response status. Specific operations include tracking the advertiser's response using a real-time monitoring system.

[1350] Step 10:

[1351] The server imposes penalties when advertisers do not take appropriate action. The input is log data on the response status, and the output is the penalty action taken. Specific actions include automatic removal of ads and application of fines.

[1352] The specific processing unit 290 transmits the result of the specific processing to the headset type terminal 314. In the headset type terminal 314, the control unit 46A causes the speaker 240 and the display 343 to output the result of the specific processing. The microphone 238 acquires audio indicating a user input regarding the result of the specific processing. The control unit 46A transmits audio data indicating the user input acquired by the microphone 238 to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the audio data.

[1353] The data generation model 58 is a so-called generative AI (Artificial Intelligence). An example of the data generation model 58 is ChatGPT (Internet Search<URL: https: / / openai.com / blog / chatgpt> ), Gemini (Internet search <url: https: gemini.google.com ?hl="ja">) and other generation AIs. The data generation model 58 is obtained by performing deep learning on a neural network. A prompt including an instruction is input to the data generation model 58, and inference data such as voice data indicating voice, text data indicating text, and image data indicating an image is also input. The data generation model 58 performs inference on the input inference data in accordance with the instruction indicated by the prompt, and outputs the inference result in a data format such as voice data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[1354] In the above embodiment, an example was given in which the specific processing is performed by the data processing device 12, but the technology of the present disclosure is not limited to this, and the specific processing may be performed by the headset type terminal 314.

[1355] [Fourth embodiment]

[1356] FIG. 7 shows an example of the configuration of a data processing system 410 according to the fourth embodiment.

[1357] 7, a data processing system 410 includes a data processing device 12 and a robot 414. An example of the data processing device 12 is a server.

[1358] The data processing device 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 is an example of a "computer" according to the technology of the present disclosure. The computer 22 includes a processor 28, a RAM 30, and a storage 32. The processor 28, the RAM 30, and the storage 32 are connected to a bus 34. The database 24 and the communication I / F 26 are also connected to the bus 34. The communication I / F 26 is connected to a network 54. Examples of the network 54 include a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[1359] The robot 414 includes a computer 36, a microphone 238, a speaker 240, a camera 42, a communication I / F 44, and a control target 443. The computer 36 includes a processor 46, a RAM 48, and a storage 50. The processor 46, the RAM 48, and the storage 50 are connected to a bus 52. The microphone 238, the speaker 240, the camera 42, and the control target 443 are also connected to the bus 52.

[1360] The microphone 238 receives instructions and the like from the user 20 by receiving voice uttered by the user 20. The microphone 238 captures the voice uttered by the user 20, converts the captured voice into audio data, and outputs it to the processor 46. The speaker 240 outputs audio in accordance with instructions from the processor 46.

[1361] Camera 42 is a small digital camera equipped with an optical system including a lens, aperture, and shutter, and an imaging element such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, and captures images of the surroundings of user 20 (for example, an imaging range defined by an angle of view equivalent to the field of vision of a typical healthy person).

[1362] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 are responsible for the exchange of various information between the processor 46 and the processor 28 via the network 54. The exchange of various information between the processor 46 and the processor 28 using the communication I / Fs 44 and 26 is carried out in a secure state.

[1363] The control object 443 includes a display device, LEDs in the eyes, and motors for driving the arms, hands, and feet. The posture and gestures of the robot 414 are controlled by controlling the motors of the arms, hands, and feet. Some of the emotions of the robot 414 can be expressed by controlling these motors. In addition, the facial expressions of the robot 414 can also be expressed by controlling the light emission state of the LEDs in the eyes of the robot 414.

[1364] Fig. 8 shows an example of the main functions of the data processing device 12 and the robot 414. As shown in Fig. 8, in the data processing device 12, a specific process is performed by the processor 28. A specific process program 56 is stored in the storage 32.

[1365] The specific processing program 56 is an example of a "program" according to the technology of the present disclosure. The processor 28 reads the specific processing program 56 from the storage 32 and executes the read specific processing program 56 on the RAM 30. The specific processing is realized by the processor 28 operating as a specific processing unit 290 in accordance with the specific processing program 56 executed on the RAM 30.

[1366] The storage 32 stores a data generation model 58 and an emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[1367] In the robot 414, the processor 46 performs the reception output process. A reception output program 60 is stored in the storage 50. The processor 46 reads the reception output program 60 from the storage 50 and executes the read reception output program 60 on the RAM 48. The reception output process is realized by the processor 46 operating as the control unit 46A in accordance with the reception output program 60 executed on the RAM 48.

[1368] Next, a description will be given of the specific processing performed by the specific processing unit 290 of the data processing device 12. In the following description, the data processing device 12 will be referred to as a "server" and the robot 414 will be referred to as a "terminal."

[1369] ---

[1370] The "Fraudulent AI Checker" system of the present invention effectively monitors advertisements and content on web services and detects false information and deep fakes generated by the use of malicious AI technology. This system is implemented primarily in a server-based configuration.

[1371] The components of this system include the following means:

[1372] 1. How you get ads and content:

[1373] The server retrieves new ads and content from the target web service using an API or scraping tool.

[1374] 2. Input to the AI ​​model:

[1375] The server converts the acquired advertisements and content into the required format and inputs them into the AI ​​model.

[1376] 3. Reliability assessment measures:

[1377] AI models analyze data within ads and content to assess their trustworthiness, covering visual, audio, and textual elements.

[1378] 4. How to identify inappropriate content:

[1379] The server identifies fraudulent advertisements and content based on the analysis results of the AI ​​model.

[1380] 5. How to request deletion or correction:

[1381] The server provides a means for sending a notice to an advertiser where fraud has been confirmed, requesting deletion or correction.

[1382] 6. How to check compatibility:

[1383] The server monitors the response status of the advertiser and checks whether the response has been made within a specified time limit.

[1384] 7. Means of imposing penalties:

[1385] To provide a means for the server to impose penalties on advertisers if they do not take appropriate action.

[1386] ---

[1387] Program processing explanation

[1388] The server first retrieves newly posted ads and content from the target web service, specifically by downloading the data using an API or scraping tool and storing it in temporary storage.

[1389] The server then converts the stored ads and content into a more easily digestible format and feeds it into an AI model that analyzes the video, audio, and text information and compares it with public sources to assess its trustworthiness.

[1390] After the AI ​​model evaluates the reliability, the server receives the analysis results and identifies fraudulent ads and content. If the reliability score falls below a certain threshold, the content is deemed fraudulent and recorded in a dedicated database.

[1391] The server will send a notice to the advertiser requesting that the advertiser remove or correct any inappropriate content that has been recorded. If the advertiser does not respond within a specified timeframe, the server will impose penalties on the advertiser, such as automatically removing the advertisement or applying a fine.

[1392] ---

[1393] Specific examples

[1394] As an example, an embodiment in a video distribution service will be described.

[1395] 1. The server calls the API to retrieve new video ads on the distribution platform.

[1396] 2. The server converts the video and audio data it receives into an analyzable format for input into the AI ​​model.

[1397] 3. An AI model analyzes data in the video ad and detects that the person in the ad was created using deepfake technology.

[1398] 4. The server determines the ad is fraudulent and sends a notification to the advertiser requesting its removal.

[1399] 5. After the server notifies the user, it checks whether the advertisement has been deleted within the specified time limit, and if the user does not comply, it automatically deletes the advertisement and imposes a penalty.

[1400] Another example is its application in blog advertising.

[1401] 1. The server retrieves newly posted ads from the blog site.

[1402] 2. The server converts the ad into text data to be input into the AI ​​model.

[1403] 3. The AI ​​model analyzes all text in the ad to detect false product information and exaggerated claims of effectiveness.

[1404] 4. The server determines that the ad is a false advertisement and sends a notice to the advertiser requesting its removal.

[1405] 5. If the advertiser does not take appropriate action after the server notifies them, the advertisements will be automatically removed from the blog site and a fine will be imposed on the advertiser.

[1406] In this way, the "Malicious AI Checker" system based on this invention provides a concrete means for quickly detecting threats posed by malicious AI use and protecting end users and web service operators.

[1407] The processing flow will be explained below.

[1408] Step 1:

[1409] The server retrieves new ads and content from the target web service. Specifically, the server uses an API or a scraping tool to download data from the web service's endpoint.

[1410] Step 2:

[1411] The server stores the acquired ads and content in temporary storage, which is fast and reliable because it is a processing stage before the data is input into the AI ​​model.

[1412] Step 3:

[1413] The server converts the stored ads and content into a more easily processed format, for example extracting video frames, transcribing audio data, or removing unnecessary tags and formatting from text data.

[1414] Step 4:

[1415] The server inputs the converted data into the AI ​​model, which then makes an API call to the AI ​​model to send the data and begin analysis.

[1416] Step 5:

[1417] AI models analyze information within ads and content, such as facial recognition and movement analysis for video, speech recognition and sentiment analysis for audio, and natural language processing for text.

[1418] Step 6:

[1419] The AI ​​model sends the analysis results to a server, which include a reliability score, suspected fraud, and specific details.

[1420] Step 7:

[1421] The server receives the analysis results from the AI ​​model and identifies fraudulent ads and content. Specifically, if the reliability score falls below a set threshold, the ad or content is deemed fraudulent.

[1422] Step 8:

[1423] The server records any advertisements or content that it determines to be fraudulent in a dedicated database, which allows it to maintain a history and be used for subsequent processing.

[1424] Step 9:

[1425] The server will then send a notice to the advertiser requesting removal or correction of the identified fraudulent content, including the reason for the fraudulent content and instructions on how to correct or remove the content.

[1426] Step 10:

[1427] The server monitors the advertiser's response status, confirms whether the advertiser has responded within the specified time limit, and records the response history.

[1428] Step 11:

[1429] If the server does not comply with the advertiser's request, it will impose penalties, such as automatic deletion of advertisements or imposition of fines.

[1430] Step 12:

[1431] The server stores all processing results and response history as logs and periodically provides reports to the web service operator, including the number of fraudulent content cases detected and the advertiser's response status.

[1432] In this way, by performing specific actions sequentially at each step, the "Fraudulent AI Checker" system effectively detects fraudulent advertisements and content and takes measures against them.

[1433] Example 1

[1434] Next, a description will be given of Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the robot 414 will be referred to as a "terminal."

[1435] Currently, advertisements and content on web services sometimes contain false information or deepfakes generated using malicious AI technology. This can lead users to make decisions based on misinformation and believe in inappropriate content or advertisements. This fraudulent content is also a major problem for web service operators, as it undermines their credibility. A system that can quickly and effectively detect and respond to these issues is needed.

[1436] The specific processing by the specific processing unit 290 of the data processing device 12 in the first embodiment is realized by the following means.

[1437] In this invention, the server includes means for acquiring advertisements and content, means for converting the acquired advertisements and content into an appropriate data format, means for inputting the converted data into a generative AI model and evaluating its reliability, means for identifying fraudulent advertisements and content based on the analysis results of the generative AI model, means for requesting deletion or correction from advertisers whose fraud has been confirmed, and means for checking the advertiser's response status and imposing penalties if the advertiser does not comply. This makes it possible to quickly detect fraudulent advertisements and content on web services and take appropriate measures.

[1438] "Advertisements and content" refers to information and media published on web services that are intended to promote purchases or provide information to users.

[1439] "Means of acquisition" refers to APIs and scraping tools used to collect advertisements and content from web services.

[1440] "Means for converting into an appropriate data format" refers to the process for converting captured advertisements or content into an analyzable format, such as image, text, audio, or video data.

[1441] A "generative AI model" is a model that uses a pre-trained artificial intelligence algorithm to evaluate the reliability and accuracy of data based on the data provided.

[1442] "Measures to assess trustworthiness" refers to the process of using generative AI models to analyze advertising and content data to determine whether the content is accurate and whether it has been generated by malicious AI technology.

[1443] "Means for identifying fraudulent advertisements and content based on analysis results" refers to the process of using the reliability assessment results output from the generative AI model to identify whether advertisements and content are fraudulent.

[1444] "Means for requesting removal or correction" refers to a notification system that requests advertisers and content providers to remove or correct advertisements or content that are deemed fraudulent.

[1445] "Measures to verify response status" refers to a process for monitoring how advertisers and content providers respond to notifications and assessing whether appropriate action has been taken within the specified timeframe.

[1446] "Measures for imposing penalties" refers to a system that implements sanctions, such as automatic removal of advertisements or imposition of fines, if advertisers or content providers do not take action within a specified deadline.

[1447] MODE FOR CARRYING OUT THE INVENTION

[1448] The "Fraudulent AI Checker" system of this invention monitors advertisements and content on web services to detect false information and deep fakes generated using malicious AI technology. This system is implemented primarily in a server-based configuration.

[1449] System Configuration

[1450] The server is the main processor and includes the following elements:

[1451] 1. How you get ads and content:

[1452] The server retrieves data from the target web service using an API or a scraping tool. An example of an API is a REST API. Examples of scraping tools include Beautiful Soup and Scrapy.

[1453] 2. Data format conversion method:

[1454] The acquired advertisements and content are converted into the appropriate data format. For example, video files are extracted as images for each frame, and audio data is converted into text. OpenCV is used for image processing, and the Google Cloud Speech-to-Text API is used for speech-to-text conversion.

[1455] 3. Input to the AI ​​model:

[1456] The server then inputs the converted data into a generative AI model, which is built using TensorFlow or PyTorch, and the model is prompted with the prompt, "Please rate the reliability of this data."

[1457] 4. Reliability assessment measures:

[1458] The generative AI model analyzes the data and assesses its reliability, specifically assessing whether the video data was generated using deepfake technology and whether the text data contains exaggeration or false information.

[1459] 5. How to identify inappropriate content:

[1460] The server receives the analysis results from the AI ​​model and identifies fraudulent ads and content based on the reliability score. If fraud is determined, the results are recorded in a dedicated database.

[1461] 6. How to request deletion or correction:

[1462] The server sends notifications to advertisers who have been found to be fraudulent, requesting them to remove or correct the content. Notifications are sent via email systems or web service notification engines.

[1463] 7. How to check for compliance:

[1464] The server monitors the advertiser's response and checks whether the appropriate action was taken within the specified time period by calling the API again and checking the data.

[1465] 8. Means of imposing penalties:

[1466] If the advertiser does not respond within the specified time frame, the server will automatically remove the ads and content and impose fines or access restrictions as necessary.

[1467] Specific examples

[1468] Examples of applications for video streaming services include:

[1469] 1. The server retrieves a new video list using the video streaming service's API.

[1470] 2. To convert the video captured by the server into an analyzable format, the video file is cut out frame by frame and the audio data is converted into text.

[1471] 3. The server inputs the converted data into the deepfake detection AI model, entering the prompt "Please rate the reliability of this data."

[1472] 4. The AI ​​model analyzes the video frames and detects that they are deepfakes.

[1473] 5. The server determines that the video is fraudulent and records it in the database.

[1474] 6. The server sends a removal request notification to the advertiser.

[1475] 7. The server checks whether the video has been deleted within the specified time limit.

[1476] 8. If the server does not respond, the video will be automatically deleted and penalties will be applied.

[1477] In this way, the "Fraudulent AI Checker" system based on this invention provides a concrete means to quickly detect the threat of false information and deep fakes caused by fraudulent AI technology and protect users and web service operators.

[1478] The flow of the identification process in the first embodiment will be described with reference to FIG.

[1479] Step 1:

[1480] Acquiring advertisements and content

[1481] The server retrieves ads and content from the target web service. Specifically, it collects data using an API or scraping tool. The retrieved data is then stored in temporary storage. The input is the web service's API endpoint or scraping tool settings, and the output is the retrieved ad and content data. This ensures that the latest ads and content are available within the system.

[1482] Step 2:

[1483] Data format conversion

[1484] The server converts the advertisements and content it acquires into an appropriate data format. Specifically, for video files, images are extracted frame by frame, and audio data is converted into text. The text information is used as is. The input is the acquired raw data, and the output is data converted into a format that can be analyzed by the generative AI model. This prepares data suitable for the AI ​​model.

[1485] Step 3:

[1486] Input to the AI ​​model

[1487] The server inputs the converted data into a generative AI model, which is built using TensorFlow or PyTorch. The model receives a prompt, "Please rate the reliability of this data." The input is the converted data and the prompt, and the output is the result of the AI ​​model's reliability assessment. This starts the process of assessing the reliability of the data.

[1488] Step 4:

[1489] Reliability assessment

[1490] The generative AI model analyzes the provided data and assesses its reliability. Specifically, it assesses whether deepfake technology has been used in video data and whether false or exaggerated information is included in text data. The input is a set of data to the AI ​​model, and the output is a reliability score or analysis results. This prepares the model for the next step based on the analysis results.

[1491] Step 5:

[1492] Identifying fraudulent content

[1493] The server receives the analysis results of the AI ​​model and identifies fraudulent ads and content based on the reliability score. A threshold score is set for what is deemed fraudulent, and content below this threshold is identified as fraudulent. The input is the reliability score and analysis results, and the output is a list of fraudulent content. This allows specific fraudulent content to be identified and addressed.

[1494] Step 6:

[1495] Requests for removal or correction

[1496] The server sends a notification to the advertiser where the fraud has been confirmed, requesting removal or correction. The notification is sent via an email system or a web service notification engine. The input is a list of fraudulent content and the advertiser's contact information, and the output is the notification sent. This allows the appropriate instructions to be given to the advertiser.

[1497] Step 7:

[1498] Check compatibility

[1499] The server monitors the advertiser's response and verifies whether appropriate action has been taken within the specified time limit. The server then calls the API again to retrieve data and verify whether the original inappropriate content has been corrected or deleted. The input is the re-retrieved data, and the output is the confirmation result of whether action has been taken. This makes it possible to understand the advertiser's response status.

[1500] Step 8:

[1501] Imposing a penalty

[1502] If the server does not take action within the specified deadline, it imposes penalties on the advertiser or content provider. Specifically, it automatically deletes the fraudulent ads and content, and imposes fines or access restrictions as necessary. The input is the response result and penalty conditions, and the output is the executed penalty. This allows inappropriate responses to be punished quickly, maintaining the reliability of the system.

[1503] (Application example 1)

[1504] Next, a description will be given of Application Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the robot 414 will be referred to as a "terminal."

[1505] In recent years, the use of malicious AI technology to create false information and deep fakes in advertisements and content has increased, increasing the likelihood of ordinary consumers being misled or harmed. There is a need for a means to quickly and effectively detect, correct, or remove such fraudulent advertisements and content. There is also a need for a system that can immediately warn viewers of the risk of fraudulent advertisements and urge them to be careful.

[1506] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 1 is realized by the following means.

[1507] In this invention, the server includes means for acquiring advertisements and content, means for inputting the acquired advertisements and content into an AI model and evaluating their reliability, means for identifying fraudulent advertisements and content based on the analysis results of the AI ​​model, means for requesting deletion or correction from advertisers whose fraud has been confirmed, means for checking the advertiser's response status and imposing penalties if the advertiser does not comply, means for automatically capturing advertisements viewed by users, means for analyzing the captured advertisements and evaluating their reliability, means for notifying users of a warning when a fraudulent advertisement is detected, and means for sending a warning to the advertiser as well. This improves the reliability of advertisements viewed by users and enables rapid detection and response to fraudulent advertisements.

[1508] "Means of obtaining advertisements and content" refers to the function of obtaining advertisements and content on web services using APIs or scraping tools.

[1509] "Means of inputting into an AI model and evaluating reliability" refers to a function that converts acquired advertisements and content into an appropriate format and inputs it into an AI model to evaluate its reliability.

[1510] "Means for identifying fraudulent advertisements and content" refers to a function that determines the reliability of advertisements and content based on the analysis results of an AI model and identifies those that contain fraudulent elements.

[1511] "Means to request removal or correction" is a function that sends a notification to the advertiser requesting removal or correction if fraudulent advertisements or content are confirmed.

[1512] "Means to check the advertiser's response status and impose penalties if they do not comply" is a function that checks whether the advertiser has responded appropriately to requests for deletion or correction, and imposes penalties if they do not comply.

[1513] "Means for automatically capturing advertisements viewed by users" refers to a function that automatically captures screenshots and URLs of advertisements viewed by users.

[1514] "Means for analyzing captured advertisements and evaluating their reliability" refers to a function that converts the captured advertisement data into an appropriate format and inputs it into an AI model to evaluate its reliability.

[1515] "Means for notifying users of a warning when fraudulent advertising is detected" is a function that displays a warning to users when fraudulent advertising is detected.

[1516] The "means for sending a warning notice to the advertiser" is a function that, when a fraudulent advertisement is detected, sends a warning notice to the advertiser that provided the advertisement.

[1517] The "Fraudulent AI Checker" system of this invention evaluates the reliability of advertisements and content viewed by users and detects false information and deep fakes. This system is implemented in a server-based configuration.

[1518] Components

[1519] 1. Server

[1520] 2. User device (smartphone)

[1521] 3. AI Model

[1522] Program structure and operation

[1523] 1. Acquiring advertisements and content

[1524] The server periodically retrieves new advertisements and content from the web service using an API or a scraping tool. The system also includes a function to automatically capture screenshots and URLs of advertisements viewed by the user's device. Smartphone screen capture APIs (such as Android's MediaProjection API and iOS's UIScreenshotService) are used for this purpose.

[1525] 2. Data conversion and input to AI models

[1526] The acquired advertisements and content are converted into an analyzable format by the server. For example, an image recognition API (such as Google Cloud Vision API) is used to convert the data into text or video data, which is then input into an AI model.

[1527] 3. Reliability evaluation

[1528] AI models (generative AI models built with TensorFlow, PyTorch, etc.) analyze ad and content data and evaluate their trustworthiness, with the aim of detecting deep fakes and false information.

[1529] 4. Identifying and warning fraudulent ads

[1530] Based on the analysis results of the AI ​​model, the server identifies fraudulent ads and content. If fraud is detected, the server displays a warning to the user. It also automatically sends a warning to the advertiser, requesting that the content be removed or corrected.

[1531] 5. Response confirmation and penalties

[1532] The server monitors whether the advertiser has responded appropriately to the deletion or correction request. If the request is not responded to within the specified time frame, the server will automatically delete the ad and impose a penalty on the advertiser. This notification is sent using services such as Firebase Cloud Messaging and SendGrid.

[1533] Specific examples

[1534] Application to video advertising

[1535] The server retrieves new video ads using the video distribution platform's API.

[1536] The acquired video advertisements are converted into a format that can be analyzed by the AI ​​model (video, audio).

[1537] AI model detects deepfakes in video ads.

[1538] The server determined the content to be fraudulent and sent a notification to the advertiser requesting its removal.

[1539] A warning message appears on the user's device stating, "This ad may be false."

[1540] If the advertiser does not comply, the server will automatically remove the video and impose a penalty.

[1541] Application to blog advertising

[1542] The server retrieves new ads from the blog site using a scraping tool.

[1543] Convert it into text data and input it into the AI ​​model.

[1544] AI models detect false product information and exaggerated claims.

[1545] The server determines the ad is fraudulent and notifies the advertiser to remove it.

[1546] A warning message appears on the user's device stating, "This ad may be false."

[1547] If advertisers do not comply, their ads will be automatically removed from blog sites and fines will be imposed.

[1548] Prompt Sentence Examples

[1549] "Enter your ad text and we'll check it for false or exaggerated information."

[1550] As described above, the system provides a concrete means to quickly assess the trustworthiness of advertisements and content, and protect users from false information and fraudulent advertisements.

[1551] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[1552] Step 1: Getting ads and content

[1553] The server periodically obtains new advertisements and content from a web service using an API or scraping tool. The input is the web service's URL or API endpoint, and the output is the newly obtained advertisement or content data. Specifically, the server calls the API and receives JSON-formatted advertisement data as a response. The server also uses a scraping tool to extract advertisement information from HTML pages and temporarily store it.

[1554] Step 2: Capture the ads users see

[1555] The device automatically captures a screenshot or URL of the ad the user is viewing. The input is the ad screen being viewed, and the output is the captured screenshot or URL. Specifically, the device's screen capture API (for example, Android's MediaProjection API or iOS's UIScreenshotService) is used to capture an image of the ad being viewed.

[1556] Step 3: Data conversion

[1557] The server converts the acquired ad and content data into an analyzable format. The input is screenshots and ad data, and the output is text or video data. Specifically, it uses an image recognition API (such as Google Cloud Vision API) to extract text from ad screenshots and convert it into the required format. In the case of video data, it uses a video analysis algorithm to split it into analyzable frames.

[1558] Step 4: Input to the AI ​​model

[1559] The server inputs the converted data into an AI model. The input is text data or video data, and the output is a reliability evaluation result. Specifically, the data is input into a generative AI model built with TensorFlow or PyTorch, and reliability evaluation is performed. The AI ​​model performs analysis based on a pre-trained dataset.

[1560] Step 5: Reliability assessment

[1561] The server receives the analysis results of the AI ​​model and evaluates the reliability of the advertisements and content. The input is the reliability score obtained from the AI ​​model, and the output is the reliability evaluation result of the advertisement or content. Specifically, if the reliability score falls below a certain threshold, it is determined to be fraudulent and recorded in a dedicated database.

[1562] Step 6: Identifying and warning fraudulent ads

[1563] The server displays a warning to the user about ads or content that is determined to be fraudulent. The input is the reliability assessment result, and the output is a warning to the user. Specifically, it uses Firebase Cloud Messaging and the device's notification function to display a message on the user's device saying, "This ad may be false."

[1564] Step 7: Notify advertisers

[1565] The server automatically sends a warning to advertisers who have been found to be fraudulent, requesting them to remove or correct the content. The input is the reliability assessment result and the advertiser's contact information, and the output is the notification sent. Specifically, an email service such as SendGrid is used to send an email to the advertiser requesting removal.

[1566] Step 8: Action Verification and Penalties

[1567] The server monitors whether the advertiser has responded appropriately to requests for deletion or correction. The input is the advertiser's response status, and the output is the monitoring results and penalty processing. Specifically, if no response is made, the server will automatically delete the advertisement and take steps to impose a penalty on the advertiser.

[1568] This process strengthens users' defenses against untrustworthy ads and encourages advertisers to take action quickly.

[1569] Furthermore, an emotion engine that estimates the user's emotion may be further combined. That is, the identification processing unit 290 may estimate the user's emotion using the emotion identification model 59, and perform identification processing using the user's emotion.

[1570] ---

[1571] By combining an emotion engine with the "Fraudulent AI Checker" system of the present invention, it is possible to monitor user reactions and further improve the reliability of advertisements and content based on the results. This system recognizes the emotions expressed by users while viewing or operating the content in real time, and reflects this data in the analysis of the AI ​​model, thereby more accurately evaluating the fraudulent nature of advertisements and content.

[1572] The components of this system include the following means:

[1573] 1. How you get ads and content:

[1574] The server retrieves new ads and content from the target web service using an API or scraping tool.

[1575] 2. Input to the AI ​​model:

[1576] The server converts the acquired advertisements and content into the required format and inputs them into the AI ​​model.

[1577] 3. Reliability assessment measures:

[1578] AI models analyze data within ads and content to assess their trustworthiness, including visual, audio, and textual elements.

[1579] 4. How to identify inappropriate content:

[1580] The server identifies fraudulent advertisements and content based on the analysis results of the AI ​​model.

[1581] 5. How to request deletion or correction:

[1582] The server provides a means for sending a notice to an advertiser where fraud has been confirmed, requesting deletion or correction.

[1583] 6. How to check compatibility:

[1584] The server monitors the response status of the advertiser and checks whether the response has been made within a specified time limit.

[1585] 7. Means of imposing penalties:

[1586] To provide a means for the server to impose penalties on advertisers if they do not take appropriate action.

[1587] 8. Emotion Engine:

[1588] The device monitors the user's reactions to what they are watching and doing in real time, collecting data. The emotion engine uses sensors such as cameras and microphones to analyze the user's facial expressions and tone of voice.

[1589] 9. Emotional data feedback methods:

[1590] The server feeds back the user's emotional data obtained from the emotion engine to the AI ​​model, allowing the user's actual reaction to be taken into account when evaluating trustworthiness.

[1591] ---

[1592] Program processing explanation

[1593] The server first retrieves newly posted ads and content from the target web service, specifically by downloading the data using an API or scraping tool and storing it in temporary storage.

[1594] The server then converts the stored ads and content into a more easily digestible format and feeds it into an AI model that analyzes the video, audio, and text information and compares it with public sources to assess its trustworthiness.

[1595] The server receives emotional data from the emotion engine on each device. When a user watches an advertisement or watches content and shows some kind of reaction to it, the emotion engine analyzes their facial expressions and tone of voice to determine their emotional state.

[1596] The AI ​​model takes emotional data into account when assessing the trustworthiness of ads and content. In addition to regular data analysis, the AI ​​model checks whether the user's emotional data indicates unnatural reactions, improving the accuracy of the assessment.

[1597] The server receives the analysis results from the AI ​​model and identifies fraudulent ads and content. If the reliability score falls below a certain threshold, the content is deemed fraudulent and recorded in a dedicated database.

[1598] The server will send a notice to the advertiser requesting that the advertiser remove or correct any inappropriate content that has been recorded. If the advertiser does not respond within a specified timeframe, the server will impose penalties on the advertiser, such as automatically removing the advertisement or applying a fine.

[1599] ---

[1600] Specific examples

[1601] As an example, an embodiment in a video distribution service will be described.

[1602] 1. The server retrieves new video ads on the distribution platform via API.

[1603] 2. The server converts the video and audio data it receives into an analyzable format for input into the AI ​​model.

[1604] 3. The user's device collects emotional data while watching the ad. The emotion engine uses facial recognition and voice analysis to analyze the user's emotional state in real time.

[1605] 4. The AI ​​model evaluates the authenticity of the video ad based on the data in the ad and the emotional data sent from the device, and if the user shows an abnormal emotional response, it determines that the ad is fraudulent.

[1606] 5. The server determines the ad is fraudulent and sends a notification to the advertiser requesting its removal.

[1607] 6. After the notification, the server will check whether the advertisement is removed within the specified time limit and monitor the compliance status. If the compliance is not met, the advertisement will be automatically removed and a fine will be imposed.

[1608] Another example is its application in blog advertising.

[1609] 1. The server retrieves newly posted ads from the blog site.

[1610] 2. The server converts the ad into text data to be input into the AI ​​model.

[1611] 3. The user's device collects emotional data while viewing the ad. The emotion engine analyzes the user's facial expressions and tone of voice to identify their emotional state.

[1612] 4. The AI ​​model analyzes all the text in the ad, takes into account emotional data, and if it finds a high likelihood of causing misunderstanding or distrust in the user, it will determine that the ad is false.

[1613] 5. The server determines that the ad is a false advertisement and sends a notice to the advertiser requesting its removal.

[1614] 6. If the advertiser does not take appropriate action after the server notifies them, the advertisements will be automatically removed from the blog site and a fine will be imposed on the advertiser.

[1615] In this way, the "Fraudulent AI Checker" system based on this invention utilizes an emotion engine to quickly and accurately detect threats posed by fraudulent AI use, providing a concrete means for protecting end users and web service operators.

[1616] The processing flow will be explained below.

[1617] Step 1:

[1618] The server obtains new ads and content from the target web service. Specifically, the server downloads the latest ad and content data from the web service endpoint using an API or scraping tool.

[1619] Step 2:

[1620] The server stores the retrieved advertisements and content in temporary storage, allowing the data to proceed to the next processing step without loss.

[1621] Step 3:

[1622] The server converts the stored ads and content into a more easily processed format, for example extracting video frames, transcribing audio data, or removing unnecessary tags and formatting from text data.

[1623] Step 4:

[1624] The server inputs the converted data into the AI ​​model, which then makes an API call to the AI ​​model to send the analyzed data, starting processing.

[1625] Step 5:

[1626] When a user's device starts displaying an advertisement or content, the device's emotion engine analyzes the user's facial expressions and voice in real time, for example, capturing the user's facial expressions with a camera and analyzing the tone of the voice with a microphone.

[1627] Step 6:

[1628] The device then sends the analyzed emotional data to the server, which includes the user's reactions to the displayed advertisements and content (such as surprise, disbelief, or anger).

[1629] Step 7:

[1630] AI models analyze data within ads and content to assess their trustworthiness, taking into account video, audio, and text information, as well as emotional data transmitted from the device.

[1631] Step 8:

[1632] The server receives the analysis results from the AI ​​model and identifies fraudulent ads and content. If the reliability score falls below a set threshold, the ad or content is deemed fraudulent.

[1633] Step 9:

[1634] The server records any ads or content that it determines to be fraudulent in a dedicated database, allowing it to maintain a history and detailed information about the fraudulent content.

[1635] Step 10:

[1636] The server will then send a notice to the advertiser requesting removal or correction of the fraudulent content, including the reason for the fraudulent content and how to correct it.

[1637] Step 11:

[1638] The server monitors the advertiser's response status, checks whether the advertiser has responded appropriately within the specified time period, and records the response details and their history.

[1639] Step 12:

[1640] The server will penalize advertisers if they fail to comply, automatically removing any abusive ads or content and notifying them of penalties if necessary.

[1641] Step 13:

[1642] The server stores all processing results and response history as logs and periodically provides reports to the web service operator, including the number of fraudulent content detected and the advertiser's response status.

[1643] ---

[1644] With this processing flow, the "Fraudulent AI Checker" system based on this invention utilizes an emotion engine to effectively identify fraudulent advertisements and content, providing a concrete means for protecting users and web service operators.

[1645] Example 2

[1646] Next, a description will be given of Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the robot 414 will be referred to as a "terminal."

[1647] The distribution of fraudulent information in modern advertising and content has become a problem. In particular, advertisements and content containing false information are rampant on the Internet, resulting in confusion for users and a loss of trust in web services. Therefore, it is necessary to quickly and accurately detect such fraudulent advertisements and content and take appropriate measures.

[1648] The identification process by the identification processing unit 290 of the data processing device 12 in Example 2 is realized by the following means. In this invention, the server includes means for acquiring advertisements and content, means for inputting the acquired advertisements and content into an AI model and evaluating their reliability, means for identifying fraudulent advertisements and content based on the analysis results of the AI ​​model, means for collecting emotional data of users viewing content on the user's terminal and feeding the data back to the AI ​​model, means for requesting deletion or correction from advertisers whose fraud has been confirmed, and means for checking the advertiser's response status and imposing penalties if the advertiser does not comply. This enables rapid and accurate detection of fraudulent advertisements and content and effective countermeasure requests to advertisers. The "advertisement and content acquisition means" refers to means used to acquire new advertisements and content from web services.

[1649] An "AI model" is an artificial intelligence algorithm used to analyze data and assess its reliability.

[1650] The "reliability evaluation means" is a means for evaluating the reliability of the advertisement or content based on the acquired advertisement or content.

[1651] "Means for identifying fraudulent advertisements and content" refers to means for identifying fraudulent advertisements and content based on the analysis results of an AI model.

[1652] The "emotion data collection means" is a means for collecting emotional data generated while a user is viewing advertisements or content.

[1653] "Emotional data feedback means" refers to a means for feeding back collected emotional data to an AI model.

[1654] "Means for requesting removal or correction" are means for requesting that advertisers whose advertisements or content have been found to be fraudulent remove or correct them.

[1655] The "compliance status confirmation means" is a means for monitoring the advertiser's compliance status.

[1656] The "penalty measure" is a measure for imposing a penalty on an advertiser if the advertiser does not take the specified action.

[1657] This invention uses a "Fraudulent AI Checker" system to monitor and evaluate the reliability of advertisements and content in real time. This system aims to more accurately evaluate the fraudulent nature of advertisements and content by recognizing the emotions expressed by users while viewing or operating the content in real time and reflecting this data in the analysis of an AI model.

[1658] System Configuration

[1659] How you get ads and content

[1660] The server accesses the target web service through an API or scraping tool to obtain new advertisements and content. For example, an API is used to obtain data from a video distribution platform, and a scraping tool is used for blog sites that contain text information.

[1661] Transforming data and inputting it into AI models

[1662] The server converts the acquired advertisements and content into an analyzable format. FFMpeg is used for video data, and Google's speech recognition system is used to convert audio data into text. Once converted, the data is formatted in JSON and input into the AI ​​model.

[1663] Reliability assessment tools

[1664] AI models analyze the video, audio, and text information in ads and content to assess their credibility, comparing it with statistical benchmarks and public databases.

[1665] Emotional data collection method

[1666] The device monitors the user's reactions in real time while watching and operating the device. Using the device's camera and microphone, the device analyzes the user's facial expressions and tone of voice to collect emotional data. For example, OpenCV and DeepFace are used to obtain facial expression data from camera footage, and the Google voice analysis system is used to obtain voice data.

[1667] Emotional Data Feedback Method

[1668] The server feeds back the user's emotional data obtained from the emotion engine to the AI ​​model, which then takes the user's emotional information into account when evaluating the reliability of the AI ​​model.

[1669] How to identify fraudulent ads and content

[1670] The server receives the analysis results from the AI ​​model and identifies fraudulent ads and content. Ads and content with a reliability score below a certain threshold are recorded as fraudulent in a dedicated database.

[1671] How to request deletion or correction

[1672] The server sends a notification to the advertiser that the fraud was detected, requesting them to remove or correct the fraud. For example, an email notification is sent using SendGrid or Mailchimp.

[1673] How to check the status of response

[1674] The server monitors the advertiser's response status and checks whether the response is made within the specified deadline. The advertiser's response status is tracked using a log monitoring system such as ElasticSearch or Kibana.

[1675] Penalty measures

[1676] The server imposes penalties if the advertiser does not take the action specified by the advertiser. It has the function to automatically delete advertisements and notify users of penalties.

[1677] Examples of concrete examples and prompts

[1678] Below are some examples and prompts:

[1679] Embodiment of video distribution service

[1680] example:

[1681] 1. The server retrieves new video ads on the distribution platform via API.

[1682] 2. The server converts the video and audio data it receives into an analyzable format for input into the AI ​​model.

[1683] 3. The user's device collects emotional data while watching the ad. The emotion engine uses facial recognition and voice analysis to analyze the user's emotional state in real time.

[1684] 4. The AI ​​model evaluates the authenticity of the video ad based on the data in the ad and the emotional data sent from the device, and if the user shows an abnormal emotional response, it determines that the ad is fraudulent.

[1685] 5. The server determines the ad is fraudulent and sends a notification to the advertiser requesting its removal.

[1686] 6. After the notification, the server will check whether the advertisement is removed within the specified time limit and monitor the compliance status. If the compliance is not met, the advertisement will be automatically removed and a fine will be imposed.

[1687] Prompt Sentence Examples

[1688] "Explain how the server retrieves new video ads from a web service and how it feeds them into the AI ​​model. Use the YouTube API as a concrete example."

[1689] "Please explain the mechanism by which the device collects emotional data from users while they are watching video ads. Please specify the sensors and analysis methods used."

[1690] Please explain in detail how your AI model uses sentiment data to evaluate the trustworthiness of ads, including the specific evaluation criteria and algorithms.

[1691] In this way, the "fraudulent AI checker" system based on this invention can achieve highly accurate monitoring and rapid response, protecting end users and web service operators.

[1692] The flow of the specific processing in the second embodiment will be described with reference to FIG. 13. Detailed Description of Processing Steps

[1693] Step 1:

[1694] The server accesses the target web service (e.g., a video streaming platform) and uses an API or scraping tool to obtain new advertising and content data.

[1695] Input: API access key for web service and scraping tool settings.

[1696] Specific behavior: The server sends a request to the specified endpoint (e.g., "https: / / example.api.com / getNewAds") and downloads the metadata and content data of the new ads.

[1697] Output: The acquired advertising and content data (e.g., video files, audio files, text data) is stored in temporary storage.

[1698] Step 2:

[1699] The server converts the acquired advertisements and content into an analyzable format and inputs it into the AI ​​model.

[1700] Input: The original data of the ads and content stored.

[1701] What it does: The server splits the video file into frames using FFMpeg and converts the audio to text, which is then formatted into JSON.

[1702] Output: Data is generated in a parsable format (JSON) that is then fed into the AI ​​model.

[1703] Step 3:

[1704] The device collects the user's reactions in real time while watching and operating the device, and analyzes emotional data.

[1705] Input: Facial and voice data of the user while watching ads and content.

[1706] Specific operation: Using the device's camera and microphone, facial expressions are analyzed using OpenCV and DeepFace, and voice tone is analyzed using the Google voice analysis system.

[1707] Output: Data representing the user's emotional state (e.g., happiness, surprise, anger, sadness, etc.) is generated and sent to the server in real time.

[1708] Step 4:

[1709] The AI ​​model assesses credibility based on data and emotional data within the ad or content.

[1710] Input: Ad / content data and user sentiment data converted into an analyzable format.

[1711] How it works: The AI ​​model analyzes video, audio, and text information, compares it to statistical benchmarks and public databases, and integrates user sentiment data to perform pattern matching and anomaly detection to detect fraud.

[1712] Output: The AI ​​model generates a credibility score for each ad or piece of content.

[1713] Step 5:

[1714] The server receives the analysis results from the AI ​​model and identifies fraudulent ads and content.

[1715] Input: Analysis results of the AI ​​model (confidence score and fraud detection information).

[1716] How it works: The server checks the reliability score and lists ads and content that fall below a certain threshold. It also records this fraudulent data in a dedicated database.

[1717] Output: A list of abusive ads and content will be generated.

[1718] Step 6:

[1719] The server sends a notice to the advertiser where fraud has been confirmed, requesting removal or correction.

[1720] Input: A list of ads and content that have been identified as fraudulent, along with details about them.

[1721] What happens next: The server uses a specialized notification system (e.g., SendGrid, Mailchimp) to send an email to the advertiser requesting removal or correction. The notification will include the specific details of the fraud and a deadline for action.

[1722] Output: A notification is sent to the advertiser.

[1723] Step 7:

[1724] The server monitors the advertiser's response status and checks whether the response is made within the specified time limit. If the appropriate response is not made, a penalty will be imposed.

[1725] Input: Advertiser response reports and log data.

[1726] Specific actions: Using log monitoring systems such as ElasticSearch and Kibana, we track the advertiser's response status. If the response is not made within the deadline, we will automatically remove the ads and apply fines.

[1727] Output: Notification of the action status report and any penalties applied, if applicable.

[1728] Through the above processing steps, the system enables rapid and accurate detection and response to fraudulent advertisements and content.

[1729] (Application example 2)

[1730] Next, a description will be given of Application Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the robot 414 will be referred to as a "terminal."

[1731] In recent years, with the increase in online advertising and content, the reliability of fraudulent advertising and content has become a problem. This problem increases the risk of users receiving misinformation, leading to a decline in trust in advertising and content providers. In such situations, there is a need for systems that can detect and respond to fraud with high accuracy and speed. However, current systems have few methods for evaluating reliability using emotional data, making it difficult to detect fraud that reflects users' actual reactions.

[1732] The identification process by the identification processing unit 290 of the data processing device 12 in Application Example 2 is realized by the following means. In this invention, the server includes: means for acquiring advertisements and content; means for inputting the acquired advertisements and content into an AI model to evaluate their reliability; means having an emotion engine for collecting user emotion data and evaluating the reliability of advertisements and content based on the collected emotion data; means for identifying fraudulent advertisements and content based on the analysis results of the AI ​​model; means for requesting deletion or correction from advertisers whose fraud has been confirmed; and means for checking the advertiser's response and imposing penalties if they do not comply. This makes it possible to utilize user emotion data to more accurately evaluate the fraudulence of advertisements and content and respond quickly.

[1733] "Means for obtaining advertisements and content" refers to the technical methods for accessing and collecting advertisements and content, specifically, the means for obtaining digital information using programs such as APIs and scraping tools.

[1734] "Means for inputting acquired advertisements and content into an AI model to evaluate its reliability" refers to a series of processes for appropriately processing acquired digital information and inputting it into an AI model to analyze the reliability of that information.

[1735] "Emotion engine for collecting user emotional data" refers to technology that uses sensors and analytical algorithms to collect a user's facial expressions, tone of voice, and other data in real time to identify the user's emotional state.

[1736] "Means for evaluating the reliability of advertisements and content based on collected emotional data" refers to a mechanism that incorporates user emotional data into the analysis results and uses that data to more accurately evaluate the reliability of advertisements and content.

[1737] "Means for identifying fraudulent advertisements and content based on the analysis results of an AI model" refers to a method for automatically identifying fraudulent or suspicious advertisements and content based on data analyzed by an AI model.

[1738] "Means for requesting removal or correction from advertisers where fraud has been confirmed" refers to a method of sending a notice to an advertiser requesting the removal or correction of content when fraud is determined to have occurred.

[1739] "Measures to check the advertiser's response and impose penalties if they do not comply" refers to methods for monitoring whether the advertiser has responded to the notice and, if they do not respond, implementing penalties such as fines or automatic removal of ads.

[1740] MODE FOR CARRYING OUT THE INVENTION

[1741] An embodiment of the present invention is a system that uses user emotion data to evaluate the reliability of advertisements and content. In this system, a server mainly plays the following roles.

[1742] System Configuration

[1743] 1. How you get ads and content:

[1744] The server uses APIs or scraping tools to retrieve newly posted ads and content, giving you access to the latest digital information.

[1745] 2. Input to the AI ​​model:

[1746] The server converts the acquired advertisements and content into an analyzable format and inputs it into an AI model, which analyzes video, audio, and text information to evaluate its reliability.

[1747] 3. Emotion Engine:

[1748] The user's device is equipped with an emotion engine that uses a camera and microphone to analyze the user's facial expressions and tone of voice in real time to collect emotion data, using image processing libraries such as OpenCV and dlib.

[1749] 4. Reliability assessment measures:

[1750] The server evaluates the reliability of advertisements and content based on AI models and user emotional data. For example, if the user's emotional response is unnatural, the server determines that the advertisement or content is fraudulent.

[1751] 5. How to identify inappropriate content:

[1752] The server identifies fraudulent advertisements and content based on the analysis results of the AI ​​model, making it possible to quickly identify unreliable digital information.

[1753] 6. How to request deletion or correction:

[1754] The server sends notifications to advertisers who have been found to be fraudulent, requesting them to remove or correct the content. Notifications are efficiently handled through an automated system.

[1755] 7. How to check for compliance:

[1756] The server monitors the advertiser's response and checks whether the response is made within the specified time limit, thereby ensuring that appropriate action is taken.

[1757] 8. Means of imposing penalties:

[1758] The server will impose penalties on advertisers if they do not take appropriate action, such as automatically removing ads or applying fines.

[1759] Specific examples

[1760] Specifically, an embodiment in a video distribution service will be described.

[1761] 1. The server retrieves new video ads on the distribution platform via API.

[1762] 2. The server converts the video and audio data it receives into an analyzable format for input into the AI ​​model.

[1763] 3. The user's device collects emotional data while watching the ad. The emotion engine uses facial recognition and voice analysis to analyze the user's emotional state in real time.

[1764] 4. The AI ​​model evaluates the authenticity of the video ad based on the data in the ad and the emotional data sent from the device, and if the user shows an abnormal emotional response, it determines that the ad is fraudulent.

[1765] 5. The server determines the ad is fraudulent and sends a notification to the advertiser requesting its removal.

[1766] 6. After the notification, the server will check whether the advertisement is removed within the specified time limit and monitor the compliance status. If the compliance is not met, the advertisement will be automatically removed and a fine will be imposed.

[1767] Example prompts to input to the generative AI model

[1768] Below is an example of a prompt sentence to input to the generative AI model.

[1769] I want to create an "Emotional Ad Checker" app that evaluates the reliability of ad content. It analyzes the user's facial expressions and voice in real time using a camera and microphone, and evaluates the reliability of the ad. The process proceeds based on the following conditions:

[1770] 1. Get advertising data from the API.

[1771] 2. Capture user data with camera and microphone.

[1772] 3. Identify emotions from facial expressions and voice.

[1773] 4. Emotional data is sent to the server to evaluate the credibility of the advertisement.

[1774] The above is a specific embodiment for carrying out the present invention.

[1775] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[1776] Step 1:

[1777] The server retrieves advertisements and content. Digital information is collected from web services using APIs or scraping tools and temporarily stored. The input is data obtained via APIs or scraping tools, and the output is advertisement and content data stored in temporary storage.

[1778] Step 2:

[1779] The advertisements and content acquired by the server are converted into an analyzable format. Video, audio, and text information are arranged into the required format. The input is the data acquired in step 1, and the output is data converted into a format suitable for the AI ​​model. Specific operations include splitting video data into frames and sampling audio data.

[1780] Step 3:

[1781] The server inputs the converted ad and content data into the AI ​​model and evaluates its reliability. The AI ​​model analyzes the data and generates a reliability score. The input is the data prepared in step 2, and the output is the reliability score and analysis results. Specifically, image recognition and text analysis are performed using a neural network.

[1782] Step 4:

[1783] The device uses an emotion engine to collect real-time facial and vocal data of users watching advertisements and content. The input is video and audio data captured by a camera and microphone, and the output is analyzed emotional data. Specific operations include facial recognition and voice tone analysis.

[1784] Step 5:

[1785] The device sends the collected emotion data to the server. The input is the analyzed emotion data, and the output is the data transmission to the server. The specific operation includes executing an HTTP request to send the emotion data to the server.

[1786] Step 6:

[1787] The server combines the analysis results of the AI ​​model with the emotional data to evaluate the trustworthiness of the advertisement or content. The input is the trustworthiness score from step 3 and the emotional data from step 5, and the output is the final trustworthiness evaluation result. Specific operations include an algorithm that integrates the emotional data and the trustworthiness score and reassess the fraud risk.

[1788] Step 7:

[1789] The server identifies fraudulent ads and content based on the results of the reliability evaluation. The input is the reliability evaluation result from step 6, and the output is a list of ads and content that are determined to be fraudulent. Specific operations include a process of comparing the evaluation result with a threshold and listing those that meet the criteria for determining fraud.

[1790] Step 8:

[1791] The server sends notifications to advertisers that have been found to be fraudulent, requesting their removal or correction. The input is a list of ads or content that have been found to be fraudulent, and the output is the notifications sent to the advertisers. Specific operations include sending notifications using an automated email system.

[1792] Step 9:

[1793] The server checks the advertiser's response status and monitors whether the response is made within the specified time limit. The input is the advertiser's response status data, and the output is a log file of the response status. Specific operations include tracking the advertiser's response using a real-time monitoring system.

[1794] Step 10:

[1795] The server imposes penalties when advertisers do not take appropriate action. The input is log data on the response status, and the output is the penalty action taken. Specific actions include automatic removal of ads and application of fines.

[1796] The specific processing unit 290 transmits the result of the specific processing to the robot 414. In the robot 414, the control unit 46A causes the speaker 240 and the control target 443 to output the result of the specific processing. The microphone 238 acquires voice indicating a user input regarding the result of the specific processing. The control unit 46A transmits voice data indicating the user input acquired by the microphone 238 to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the voice data.

[1797] The data generation model 58 is a so-called generative AI (Artificial Intelligence). An example of the data generation model 58 is ChatGPT (Internet Search<URL: https: / / openai.com / blog / chatgpt> ), Gemini (Internet search <url: https: gemini.google.com ?hl="ja">) and other generation AIs. The data generation model 58 is obtained by performing deep learning on a neural network. A prompt including an instruction is input to the data generation model 58, and inference data such as voice data indicating voice, text data indicating text, and image data indicating an image is also input. The data generation model 58 performs inference on the input inference data in accordance with the instruction indicated by the prompt, and outputs the inference result in a data format such as voice data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[1798] In the above embodiment, an example was given in which the specific processing is performed by the data processing device 12, but the technology of the present disclosure is not limited to this, and the specific processing may be performed by the robot 414.

[1799] The emotion identification model 59 as an emotion engine may determine the user's emotion according to a specific mapping. Specifically, the emotion identification model 59 may determine the user's emotion according to an emotion map (see FIG. 9), which is a specific mapping. Similarly, the emotion identification model 59 may determine the robot's emotion, and the identification processing unit 290 may perform identification processing using the robot's emotion.

[1800] FIG. 9 is a diagram illustrating an emotion map 400 on which multiple emotions are mapped. In the emotion map 400, emotions are arranged in concentric circles radiating from the center. Emotions closer to the center of the concentric circles are more primitive. Emotions representing states and actions arising from a state of mind are arranged on the outer edges of the concentric circles. The concept of emotion includes both affect and mental states. Emotions generally generated from reactions occurring in the brain are arranged on the left side of the concentric circles. Emotions generally induced by situational judgment are arranged on the right side of the concentric circles. Emotions generally generated from reactions occurring in the brain and induced by situational judgment are arranged on the upper and lower sides of the concentric circles. Furthermore, the emotion of "pleasure" is arranged on the upper side of the concentric circles, and the emotion of "discomfort" is arranged on the lower side. In this way, in the emotion map 400, multiple emotions are mapped based on the structure by which emotions are generated, and emotions that tend to occur simultaneously are mapped close to each other.

[1801] These emotions are distributed in the 3 o'clock direction on emotion map 400, and typically fluctuate between relief and anxiety. In the right half of emotion map 400, situational awareness dominates over internal sensations, resulting in a sense of calm.

[1802] The inside of emotion map 400 represents what is going on in the mind, and the outside of emotion map 400 represents behavior, so the further you go outside emotion map 400, the more visible the emotions become (the more they are expressed in behavior).

[1803] Human emotions are based on various balances, such as posture and blood sugar levels. When these balances deviate from the ideal, a state of discomfort is indicated, and when they approach the ideal, a state of pleasure is indicated. Emotions can also be created for robots, automobiles, and motorcycles, based on various balances, such as posture and remaining battery life. When these balances deviate from the ideal, a state of discomfort is indicated, and when they approach the ideal, a state of pleasure is indicated. An emotion map can be generated, for example, based on Dr. Mitsuyoshi's emotion map (Research on Voice Emotion Recognition and Emotional Brain Physiological Signal Analysis Systems, Tokushima University, Doctoral Dissertation: https: / / ci.nii.ac.jp / naid / 500000375379). The left half of the emotion map lists emotions belonging to the "reaction" domain, where sensation is dominant. The right half of the emotion map lists emotions belonging to the "situation" domain, where situational awareness is dominant.

[1804] The emotion map defines two emotions that promote learning. One is a negative emotion on the situation side, around the middle of "repentance" or "reflection." In other words, this occurs when the robot experiences negative emotions such as "I never want to feel this way again" or "I don't want to be scolded again." The other is a positive emotion on the response side, around "desire." In other words, this occurs when the robot experiences positive feelings such as "I want more" or "I want to know more."

[1805] The emotion identification model 59 inputs user input into a pre-trained neural network, obtains emotion values ​​indicating each emotion shown in the emotion map 400, and determines the user's emotion. This neural network is pre-trained based on multiple pieces of training data that are combinations of user input and emotion values ​​indicating each emotion shown in the emotion map 400. Furthermore, this neural network is trained so that emotions that are located close to each other have similar values, as in the emotion map 900 shown in FIG. 10. FIG. 10 shows an example in which multiple emotions, "relieved," "calm," and "reassuring," have similar emotion values.

[1806] The system according to the present disclosure has been described above mainly with respect to the functions of the data processing device 12, but the system according to the present disclosure is not necessarily implemented on a server. The system according to the present disclosure may be implemented as a general information processing system. The present disclosure may be implemented, for example, as a software program running on a personal computer or an application running on a smartphone, etc. The method according to the present disclosure may be provided to users in the form of SaaS (Software as a Service).

[1807] In the above embodiment, an example was given in which the specific processing is performed by one computer 22, but the technology of the present disclosure is not limited to this, and the specific processing may be distributed and performed by a plurality of computers including the computer 22. For example, the data generation model 58 may be provided in an external device of the data processing device 12, and data may be generated in the external device in accordance with input data.

[1808] In the above embodiment, an example in which the specific processing program 56 is stored in the storage 32 has been described, but the technology of the present disclosure is not limited to this. For example, the specific processing program 56 may be stored in a portable, computer-readable, non-transitory storage medium such as a USB (Universal Serial Bus) memory. The specific processing program 56 stored in the non-transitory storage medium is installed in the computer 22 of the data processing device 12. The processor 28 executes the specific processing in accordance with the specific processing program 56.

[1809] Alternatively, the specific processing program 56 may be stored in a storage device such as a server connec...

Claims

1. a means for obtaining advertising and content; A means for inputting the acquired advertisements and content into an AI model to evaluate their trustworthiness; and A method for identifying fraudulent ads and content based on the analysis results of the AI ​​model; A means to request removal or correction from advertisers where fraud has been identified; A system that includes a means to check advertiser compliance and impose penalties for non-compliance.

2. The system according to claim 1, wherein the means for obtaining advertisements and content uses an API or a scraping tool.

3. 2. The system of claim 1, wherein the means for requesting deletion or modification includes a means for sending a notification.

Images