system

Abstract

A system is provided.SOLUTION: A system comprising: means for receiving email messages and social networking service messages; means for analyzing the received messages to detect potentially fraudulent text and unreliable URLs; means for displaying an alert in the event of potential fraud; means for automated voice response to incoming calls from anonymous or unknown phone numbers and analyzing call content in real-time; and means for updating models to learn past fraud cases and respond to new fraud techniques.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] In modern society, frauds are becoming increasingly sophisticated, with an increasing number of frauds using email, social networking services, and telephone calls. This has resulted in frequent leaks of personal information and financial losses. Conventional countermeasures cannot completely prevent these frauds, and there is a problem that elderly people and those unfamiliar with using the Internet are particularly susceptible to falling victim to them. The aim of this project is to solve these problems and reduce the damage caused by fraud. [Means for solving the problem]

[0005] The present invention solves the above-mentioned problems with a system that includes a means for receiving email messages and messages from social networking services, a means for analyzing received messages to detect potentially fraudulent wording and unreliable URLs, a means for displaying an alert when there is a possibility of fraud, a means for responding to calls from blocked or unknown phone numbers with an automated voice and analyzing the content of the call in real time, and a means for learning from past fraud cases and updating a model to adapt to new fraud methods. Specifically, the system uses a natural language processing algorithm on the content of received messages and calls to evaluate the possibility of fraud, and warns the user through an alert or call confirmation as necessary, thereby preventing fraudulent acts and reducing damage.

[0006] An "email message" is a text-based communication sent or received over the Internet.

[0007] A "social networking service" is a platform through which people communicate online.

[0008] An "alert" is a notification or warning message intended to alert the user.

[0009] An "anonymous call" is a call made in such a way that the caller's phone number is not displayed to the recipient.

[0010] An "unknown phone number" is a phone number that the recipient has not previously contacted or has no caller information on file.

[0011] "Automatic voice response" is a system that automatically responds to callers using AI or pre-recorded messages.

[0012] "Call content analysis" is the process of converting the content of a telephone conversation into text in real time and analyzing the meaning based on that information.

[0013] "Possibility of fraud" is a concept that refers to the degree to which a particular act or word is suspected to be fraudulent.

[0014] A "natural language processing algorithm" is an algorithm that performs the computational tasks of understanding, interpreting, and generating human language.

[0015] "Model updating" is the process by which a machine learning data analysis system improves itself based on new data and information. [Brief explanation of the drawings]

[0016] [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

[0017] 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.

[0018] First, the terms used in the following description will be explained.

[0019] 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).

[0020] 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.

[0021] 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.

[0022] 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.

[0023] 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."

[0024] [First embodiment]

[0025] FIG. 1 shows an example of the configuration of a data processing system 10 according to the first embodiment.

[0026] 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.

[0027] 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).

[0028] 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.

[0029] 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.

[0030] 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.

[0031] 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.

[0032] FIG. 2 shows an example of the main functions of the data processing device 12 and the smart device 14.

[0033] 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.

[0034] 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.

[0035] 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.

[0036] 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."

[0037] The present invention relates to a system for reducing the risk of special fraud occurring through email, social networking services, and telephone. This system is realized by a program having the following functions.

[0038] Email and SNS fraud detection function

[0039] When a device receives a new email or message from a social networking service, the server analyzes the message's content. Specifically, the server uses natural language processing algorithms to analyze the message text and detect potentially fraudulent phrases and unreliable URLs. If the server determines that a message is potentially fraudulent, it displays an alert message on the device to warn the user.

[0040] Examples:

[0041] 1. Your device will receive an email asking you to enter your bank account information.

[0042] 2. The server analyzes the text and detects possible fraud.

[0043] 3. The device notifies the user, "This email may be fraudulent. Please be careful."

[0044] Telephone fraud response features

[0045] When the device receives a call from a blocked or unknown number, the device will initiate an automated voice response. AI will analyze the call content in real time, and if there is a possibility of fraud, the server will notify the user of the call content and request confirmation of the call.

[0046] Examples:

[0047] 1. The device receives an anonymous call.

[0048] 2. The device will respond with an automated voice saying, "This call is being recorded for security reasons. Please wait as this may be a scam."

[0049] 3. The AI ​​analyzes the content of the call and detects the phrase "I urgently need money."

[0050] 4. The server detects possible fraud and

[0051] 5. The device notifies the user, "You have received an anonymous call. Please contact your family or the police to confirm the details."

[0052] Fraudulent wording learning function

[0053] The server collects data on past fraud cases and uses machine learning algorithms to learn fraudulent phrases. This allows the model to be updated to handle new fraud methods and enable early detection of unknown fraud methods. If a new fraud method is detected, the device will display a warning to the user based on that information.

[0054] Examples:

[0055] 1. The server analyzes past fraudulent emails and learns the common phrase "Please transfer the money quickly."

[0056] 2. The server adds the new fraud technique to the learning model.

[0057] 3. The server detects a new fraudulent email that says, "Click this link to receive a special discount."

[0058] 4. The device notifies the user, "This email may be a new scam. Do not click on the link."

[0059] As described above, the system according to the present invention can detect fraudulent acts via email, social media, and telephone, and can prevent fraud damage by issuing a warning to the user in advance, thereby effectively reducing fraud damage.

[0060] The processing flow will be explained below.

[0061] Email and SNS fraud detection function

[0062] Step 1:

[0063] Your device receives a new email or SNS message.

[0064] The device will notify you when a new message arrives in your inbox.

[0065] Step 2:

[0066] Gets the contents of the message received by the server.

[0067] The server stores the message data sent from the terminal in an analysis buffer.

[0068] Step 3:

[0069] The server begins parsing the message.

[0070] The server uses natural language processing (NLP) algorithms to tokenize and pattern match the message text.

[0071] Step 4:

[0072] The server assesses the likelihood of fraud.

[0073] The server determines the likelihood that the message content is fraudulent based on a registered list of fraudulent phrases and a reliability score.

[0074] Step 5:

[0075] Determines whether the server generates an alert.

[0076] The server decides to generate an alert message if the likelihood of fraud exceeds a threshold.

[0077] Step 6:

[0078] The device displays an alert to the user.

[0079] The device will display a pop-up warning message saying, "This email may be fraudulent. Please be careful."

[0080] Telephone fraud response features

[0081] Step 1:

[0082] Your device receives a call from a blocked or unknown number.

[0083] The device retrieves the caller information and matches it with an existing contact database.

[0084] Step 2:

[0085] The device will start an automatic voice response.

[0086] The terminal will begin the process of playing a pre-configured auto-answer message.

[0087] Step 3:

[0088] AI analyzes the content of calls in real time.

[0089] AI uses voice recognition technology to convert calls into text data, which is then analyzed using NLP algorithms.

[0090] Step 4:

[0091] The server assesses the likelihood of fraud.

[0092] Based on the analysis results, the server scores the likelihood that the call content is fraudulent.

[0093] Step 5:

[0094] The server asks the user to confirm the call.

[0095] If the server determines that there is a high possibility of fraud, it sends a warning message to the user's terminal.

[0096] Step 6:

[0097] The terminal displays a call confirmation to the user.

[0098] The device will display a message saying, "You have received an anonymous call. Please contact your family or the police to confirm the details."

[0099] Fraudulent wording learning function

[0100] Step 1:

[0101] The server collects historical fraud data.

[0102] The server stores past fraudulent emails and phone call details in a database.

[0103] Step 2:

[0104] The server learns the fraudulent language.

[0105] The server uses machine learning algorithms to analyze and extract common fraud patterns from the collected data.

[0106] Step 3:

[0107] The server updates the model.

[0108] The server adds and updates the learning model with newly discovered fraud techniques and patterns.

[0109] Step 4:

[0110] The server detects new fraud methods in real time.

[0111] The server applies the updated model to detect new fraud techniques from the latest data.

[0112] Step 5:

[0113] The terminal alerts the user based on new fraud patterns.

[0114] The device will display a notification saying, "This email may be a new scam. Do not click on any links."

[0115] The above are the processing steps in a specific embodiment of the present invention.

[0116] Example 1

[0117] 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."

[0118] Conventional systems have had difficulty effectively reducing the risk of fraud via email, social media, and telephone. In particular, they were unable to adapt to new fraud methods, increasing the likelihood of fraud victims occurring. Furthermore, users often felt uneasy because they were unable to analyze call content in real time to detect potential fraud.

[0119] 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.

[0120] In this invention, the server includes: means for receiving email messages and messages from social networking services; means for analyzing the received messages to detect potentially fraudulent wording and unreliable URLs; means for displaying an alert when there is a possibility of fraud; means for responding to calls from anonymous or unknown phone numbers with an automated voice and analyzing the content of the call in real time; means for learning from past fraud cases and updating the model to respond to new fraud techniques; and means for displaying a warning to the user when there is a possibility of fraud among the above means; means for analyzing the content of messages and calls in real time using a natural language processing algorithm to detect possible fraud; and means for detecting new fraud techniques using a generative AI model based on the analysis results. This makes it possible to detect fraudulent acts via email, social networking services, and telephone calls early and prevent fraud damage by issuing advance warnings to users.

[0121] "Email message" refers to the text information of an email sent or received over the Internet.

[0122] A "social networking service" is a platform that allows people to share information and communicate online.

[0123] "Means of receiving" refers to the technology and equipment used to receive email messages and messages from social networking services.

[0124] "Means for analysis" refers to the technology and equipment used to analyze the content of received messages and find specific patterns or characteristics.

[0125] "Potentially fraudulent language and unreliable URLs" refers to keywords associated with fraudulent activities and unreliable web links.

[0126] "Means for displaying an alert" refers to the technology and devices used to display a warning message to the user when a potential fraud is detected.

[0127] "Anonymous or unknown phone numbers" are phone numbers that cannot be identified by the caller and that have not been previously registered.

[0128] "Automatic voice response means" means technology and equipment that automatically plays a voice message in response to an incoming call without human intervention.

[0129] "Means for analyzing call content in real time" refers to technology and equipment that allows for immediate analysis of the content of a call while it is being made.

[0130] "Past fraud cases" refers to specific cases and data of fraudulent acts that have occurred in the past.

[0131] "Model updating methods" refers to the techniques and methods used to keep fraud detection algorithms up to date based on new information and data.

[0132] A "natural language processing algorithm" is an algorithm that allows a computer to analyze and understand human language.

[0133] A "generative AI model" is an artificial intelligence model that uses machine learning techniques to generate new patterns and knowledge from data.

[0134] "Means for displaying a warning to the user" refers to technologies and devices that provide visual or audio warnings to the user when possible fraud is detected.

[0135] The present invention relates to a system for reducing the risk of special frauds occurring through email, social networking services (SNS), and telephone. This system includes functions for receiving email and SNS messages, analyzing the received messages, determining the possibility of fraud, displaying alerts, responding to telephone fraud, and learning fraudulent phrases.

[0136] Email and SNS fraud detection function

[0137] The server receives emails and SNS messages using an internet-connected device (smartphone, PC, tablet, etc.). The received message is sent to the server via a secure communication protocol (e.g., HTTPS). The server analyzes the content of the message using a natural language processing algorithm (e.g., spaCy, NLTK) to detect specific keywords (e.g., 'bank account information', 'transfer') and unreliable URLs. If it determines that there is a possibility of fraud, an alert message is generated and sent to the device. The device notifies the user, "This email may be fraudulent. Please be careful."

[0138] Specific examples

[0139] When the terminal receives an email saying "Please enter your bank account information," it sends the received information to the server.

[0140] The server analyzes the email and detects possible fraud.

[0141] The device will notify the user, "This email may be fraudulent. Please be careful."

[0142] Prompt Sentence Examples

[0143] Here's a fraud warning message for when you receive an email that says "enter your bank account details."

[0144] Telephone fraud response features

[0145] When the device receives a call from an anonymous or unknown phone number, an automated voice response begins, saying, "This call is being recorded for safety reasons. Please wait as this may be a scam." The content of the call is analyzed in real time by AI, and the results are sent to the server. The server analyzes the call text, and if it detects phrases that are likely to be fraudulent (e.g., 'I urgently need money'), it sends a notification to the user requesting that they check the content of the call. The device notifies the user, "You have received an anonymous call. Please contact your family or the police to confirm the content."

[0146] Specific examples

[0147] When the device receives an anonymous call, an automated voice responds saying, "This call is being recorded for safety reasons. Please wait as this may be a scam."

[0148] AI analyzes the content of calls in real time and detects phrases such as "I urgently need money."

[0149] The server determines the possibility of fraud, and the device notifies the user, "You have received an anonymous call. Please contact your family or the police to confirm the details."

[0150] Prompt Sentence Examples

[0151] Display a warning message if an anonymous caller claims to be in urgent need of money.

[0152] Fraudulent wording learning function

[0153] The server collects data on past fraud cases and analyzes it using machine learning algorithms (e.g., TensorFlow, Scikit-learn). It learns characteristic patterns and keywords from the analyzed data and updates the model to respond to new fraud methods. The server detects new fraud methods and sends a notification to the device. The device displays a warning to the user saying, "This email may be a new scam. Do not click on the link."

[0154] Specific examples

[0155] The server analyzes past fraudulent emails and learns the phrase "Please transfer the money quickly."

[0156] The server adds new fraud techniques to the learning model.

[0157] The server now detects the phrase "Click this link to receive a special discount."

[0158] The device will notify the user, "This email may be a new scam. Do not click on the link."

[0159] Prompt Sentence Examples

[0160] A fraud warning message for those who receive emails that say, "Click this link to receive a special discount."

[0161] This system can detect fraudulent activities via email, social media, and telephone at an early stage and warn users in advance, preventing fraud damage before it occurs. This will significantly reduce the damage caused by fraud.

[0162] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0163] Email and SNS fraud detection function

[0164] Processing Steps

[0165] Step 1:

[0166] The device receives new emails or SNS messages and saves their contents to local storage. The input is the received message, and the output is the saved message data.

[0167] Step 2:

[0168] The terminal sends the received message to the server using a secure communication protocol (e.g. HTTPS). The input is the stored message data, and the output is the message data sent to the server.

[0169] Step 3:

[0170] The server analyzes the message data it receives. Specifically, it uses a natural language processing library (e.g., spaCy, NLTK) to tokenize the message, tag parts of speech, and analyze context. The input is the message data sent to the server, and the output is the analysis results.

[0171] Step 4:

[0172] The server determines the likelihood of fraud based on the analysis results. Criteria for determining fraud include specific keywords (e.g., 'bank account information', 'transfer') and untrustworthy URLs. The input is the analysis results, and the output is an assessment of the likelihood of fraud.

[0173] Step 5:

[0174] If the server detects a possibility of fraud, it generates an alert message and sends it to the terminal. The input is the evaluation result of the possibility of fraud, and the output is the generated alert message.

[0175] Step 6:

[0176] The terminal notifies the user of the alert message. The input is the generated alert message, and the output is the warning message displayed to the user.

[0177] Telephone fraud response features

[0178] Processing Steps

[0179] Step 1:

[0180] The terminal receives a call from a blocked or unknown phone number. The input is the call from the blocked or unknown phone number, and the output is the acceptance of the call.

[0181] Step 2:

[0182] The terminal starts an automated voice response, saying, "This call is being recorded for security reasons. Please wait as this may be a scam." The input is the acceptance of the call, and the output is the automated voice response.

[0183] Step 3:

[0184] AI analyzes the contents of calls in real time. Specifically, it converts the contents of calls into text using the Google Speech-to-Text API and performs keyword analysis. The input is the contents of the calls, and the output is the analysis results.

[0185] Step 4:

[0186] The server determines the likelihood of fraud based on the analysis results. Fraud criteria include specific phrases (e.g., 'I urgently need money'). The input is the analysis results, and the output is an assessment of the likelihood of fraud.

[0187] Step 5:

[0188] If the server detects the possibility of fraud, it generates a notification requesting the user to confirm the contents of the call and sends it to the terminal. The input is the result of the evaluation of the possibility of fraud, and the output is the generated notification.

[0189] Step 6:

[0190] The device sends a notification to the user, warning them that "An anonymous call has been made. Please contact your family or the police to confirm the details." The input is the generated notification, and the output is the warning message that is displayed to the user.

[0191] Fraudulent wording learning function

[0192] Processing Steps

[0193] Step 1:

[0194] The server collects past fraudulent email and call data from the Internet and internal databases. The input is past fraud case data, and the output is the collected data.

[0195] Step 2:

[0196] The server uses machine learning algorithms (e.g., TensorFlow, Scikit-learn) to analyze the collected data and learn specific patterns or keywords. The input is the collected data, and the output is an updated learning model.

[0197] Step 3:

[0198] The server adds new fraud techniques to the learning model. The input is specific patterns or keywords, and the output is an updated learning model.

[0199] Step 4:

[0200] The server analyzes newly received emails and messages to detect new fraud methods. The input is the new email or message, and the output is the evaluation result of the likelihood of fraud.

[0201] Step 5:

[0202] If the device detects a possible scam, it notifies the user, "This email contains a new potential scam. Do not click on the link." The input is the result of the evaluation of the possible scam, and the output is a warning message that is displayed to the user.

[0203] (Application example 1)

[0204] 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."

[0205] In modern society, fraudulent activities via email, social networking services (SNS), and telephone are on the rise, and many people are falling victim to these scams. Elderly people and those unfamiliar with the Internet have particular difficulty distinguishing between fraudulent messages and phone calls, making it easier for the damage to spread. There is a need for technology that can effectively detect such fraudulent activities and issue early warnings to users.

[0206] 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.

[0207] In this invention, the server includes means for receiving email messages and messages from social networking services, means for analyzing the received messages to detect potentially fraudulent wording and unreliable URLs, means for displaying an alert when there is a possibility of fraud, means for responding to calls from withheld or unknown phone numbers with an automated voice and analyzing the content of the call in real time, means for learning from past fraud cases and updating the model to deal with new fraud methods, and means for visually warning users of the risk of fraud using a display device worn by the user. This makes it easier for users to recognize the risk of fraud in real time and prevents them from becoming victims of fraud.

[0208] "Mail message" refers to information, including text and attachments, sent and received via email.

[0209] A "social networking service" is an online platform that enables people to share information and deepen their interactions via the Internet.

[0210] A "server" is a computer system that processes and provides data over a network.

[0211] "Means for receiving messages" refers to technology that has the function of receiving emails and SNS messages on a user's device.

[0212] "Means for analyzing" refers to technology for analyzing the content of received messages to assess fraud risk.

[0213] "Wording" refers to a sentence or phrase, and refers to specific text that may be fraudulent.

[0214] "URL" stands for Uniform Resource Locator and is a description that indicates the address to a web page or online resource.

[0215] "Means for displaying an alert" refers to technology that displays a warning or alert message to the user.

[0216] "Calls from blocked or unknown numbers" refers to calls from phones that do not display or are not registered and cannot be identified.

[0217] "Means for responding with automated voice" refers to technology in which a system automatically generates and responds with a voice message.

[0218] "Means for analyzing call content in real time" refers to technology that instantly converts voice data during a call into text, analyzes the content, and assesses the risk of fraud.

[0219] "Past fraud cases" refers to data on fraud methods and cases that have been reported to date.

[0220] "Means of learning and updating models to adapt to new fraud techniques" refers to a technique that uses machine learning algorithms to learn new fraud patterns and update existing detection models.

[0221] A "display device" refers to hardware for visually presenting information, such as smart contact lenses or smart glasses.

[0222] "Means for visually warning of fraud risk" refers to technology that uses a visual display device worn by the user to display a warning about fraud risk.

[0223] The present invention relates to a system for reducing the risk of special frauds occurring through email, social networking services (SNS), and telephone. This system is realized by having the following main functions.

[0224] 1. Email and SNS fraud detection function

[0225] When a device receives a new email or message from a social networking service, the server analyzes the message's content. Specifically, the server uses a natural language processing library (e.g., Spacy) and a machine learning model (e.g., a model saved by scikit-learn) to analyze the message text and detect potentially fraudulent phrases or unreliable URLs. If any are detected, a warning message is displayed on the device in real time to alert the user.

[0226] 2. Telephone fraud response function

[0227] When the device receives a call from a blocked or unknown number, the device will initiate an automated voice response. AI will analyze the call content in real time, and if there is a possibility of fraud, the server will notify the user of the call content and request confirmation of the call.

[0228] 3. Fraudulent wording learning function

[0229] The server collects data on past fraud cases and uses machine learning algorithms to learn fraudulent phrases. This allows the model to be updated to handle new fraud methods and enable early detection of unknown fraud methods. If a new fraud method is detected, the device will display a warning to the user based on that information.

[0230] 4. Warning display function on smart contact lenses

[0231] The system also includes a function that visually warns users of fraud risks using display devices such as smart contact lenses or smart glasses. If the server detects a risk of fraud while the user is checking email or social media messages, a visual warning message saying "Possible fraud" will be displayed on the display device.

[0232] Hardware / software configuration used

[0233] Server: Analyzes messages, runs AI models, and updates learning models.

[0234] Device: Receives emails and SNS messages, records phone calls, and provides automated voice responses.

[0235] Natural language processing libraries such as Spacy can be used to analyze message content.

[0236] Machine learning models: Assess fraud risk using models saved in Scikit-learn.

[0237] Display devices: Use smart contact lenses or smart glasses to display warning messages.

[0238] Specific examples

[0239] For example, if a user receives an email requesting "Please enter your bank account information," the server analyzes the text and detects the possibility of fraud. The device (such as a smartphone) then displays a warning message saying, "This email may be fraudulent. Please be careful." At the same time, if the user is wearing smart contact lenses, the lenses will also display a visual warning saying, "This email may be fraudulent."

[0240] Prompt Sentence Examples

[0241] An example prompt for new messages would look something like this:

[0242] "A new message has arrived. Please analyze its contents."

[0243] As a result, this system can effectively detect fraudulent activities via email, social media, and telephone, and can prevent fraud damage by issuing advance warnings to users, thereby effectively reducing fraud damage.

[0244] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0245] Step 1:

[0246] The device receives a new email or message from a social networking service. The received message data is provided as input. The device sends this message to the server.

[0247] Step 2:

[0248] The server analyzes the content of the received message. The input is the text data of the received message, and the output is the analysis result. The server uses a natural language processing library (e.g., Spacy) to tokenize the message text and detect potentially fraudulent phrases and unreliable URLs.

[0249] Step 3:

[0250] The server determines the likelihood of fraud based on the analysis results. The input is the message analysis results, and the output is the fraud risk assessment result. The fraud risk is assessed using a machine learning model trained on past fraud cases.

[0251] Step 4:

[0252] If there is a possibility of fraud, the server sends a warning to the terminal and display device. The input is the fraud risk assessment result (if the fraud risk is high), and the output is a warning message to be displayed to the user. The server generates a warning message such as "Possible fraud" and sends it to the terminal and display device such as a smart contact lens.

[0253] Step 5:

[0254] When a user receives a call from a blocked or unknown phone number, the device starts an automated voice response. The input is the incoming call information, and the output is an automated voice message. The device generates an automated voice message informing the user, "This call is being recorded for safety reasons. Please wait as this may be a scam."

[0255] Step 6:

[0256] The server analyzes the call content in real time. The input is the voice data of the call, and the output is the text conversion and analysis results of the call content. The server uses AI to convert the call content into text and analyze it.

[0257] Step 7:

[0258] The server evaluates the possibility of fraud based on the analysis results of the call content. The input is the text data of the call content and the analysis results, and the output is the fraud risk assessment result.

[0259] Step 8:

[0260] If there is a possibility of fraud, the server sends a warning to the terminal and display device and requests the user to confirm the call. The input is the fraud risk assessment result (if the fraud risk is high), and the output is a warning message to be displayed to the user and a request to confirm the content of the call. The server generates a warning message saying "There has been an anonymous call. Please consult with your family or the police to confirm the content," and sends it to the terminal and display device.

[0261] Step 9:

[0262] The server collects data on past fraud cases and uses a machine learning algorithm to learn fraudulent phrases. The input is data on past fraud cases, and the output is an updated fraud detection model. If a new fraud technique is detected, it is added to the learning model and reflected in the next analysis.

[0263] Through the above processing steps, this system can effectively detect fraudulent activities via email, social media, and telephone, and can prevent fraudulent acts by issuing advance warnings to users.

[0264] 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.

[0265] The present invention combines an emotion engine with a system for reducing the risk of special frauds occurring through email, social networking services, and telephone calls, enabling appropriate responses to be taken according to the user's emotional state. This system is realized by a program with the following functions:

[0266] Email and SNS fraud detection function

[0267] When a device receives a new email or message from a social networking service, the server retrieves and analyzes the message content. Specifically, the server uses a natural language processing (NLP) algorithm to analyze the message text and detect potentially fraudulent language or unreliable URLs. If the possibility of fraud is assessed, an alert message is displayed on the device to warn the user. At this time, the emotion engine recognizes the user's emotional state, and if the user is in a state of stress, a stronger alert is displayed.

[0268] Examples:

[0269] 1. Your device will receive an email asking you to enter your bank account information.

[0270] 2. The server analyzes the text and detects possible fraud.

[0271] 3. The emotion engine analyzes the user's emotional state and sets the appropriate alert intensity.

[0272] 4. The device will pop up a warning message saying, "This email may be fraudulent. Please be careful."

[0273] Telephone fraud response features

[0274] When a device receives a call from a blocked or unknown number, it will initiate an automated voice response. AI analyzes the call content in real time, and if there is a possibility of fraud, the server will notify the user of the call content and request confirmation. During this process, an emotion engine analyzes the user's emotional state in real time and adjusts the intensity and timing of the warning based on emotional changes.

[0275] Examples:

[0276] 1. The device receives an anonymous call.

[0277] 2. The device will respond with an automated voice saying, "This call is being recorded for security reasons. Please wait as this may be a scam."

[0278] 3. The AI ​​analyzes the content of the call and detects the phrase "I urgently need money."

[0279] 4. The emotion engine analyzes the user's emotional state during the call, and if the user feels anxious, the server generates an alert with appropriate intensity.

[0280] 5. The device will display the message, "You have received an anonymous call. Please contact your family or the police to confirm the details."

[0281] Fraudulent wording learning function

[0282] The server collects data on past fraud cases and uses machine learning algorithms to learn fraudulent phrases. This allows the model to be updated to handle new fraud methods, enabling early detection of unknown fraud methods. The emotion engine also learns from the user's emotional data, allowing it to understand the user's unique emotional patterns and more precisely determine the likelihood of fraud. If a new fraud method is detected, the device will display a warning to the user based on that information.

[0283] Examples:

[0284] 1. The server analyzes past fraudulent emails and learns the common phrase "Please transfer the money quickly."

[0285] 2. The server adds the new fraud technique to the learning model.

[0286] 3. The server detects a new fraudulent email that says, "Click this link to receive a special discount."

[0287] 4. The emotion engine uses user emotional data to more precisely assess the likelihood of fraud.

[0288] 5. Your device will display a notification saying, "This email may be a new scam. Do not click on any links."

[0289] As described above, the system according to the present invention can detect fraudulent acts via email, social media, and telephone, and take appropriate measures according to the user's emotional state to prevent fraud damage before it occurs, thereby effectively reducing fraud damage.

[0290] The processing flow will be explained below.

[0291] Email and SNS fraud detection function

[0292] Step 1:

[0293] Your device receives a new email or SNS message.

[0294] The device will notify you when a new message arrives in your inbox.

[0295] Step 2:

[0296] Gets the contents of the message received by the server.

[0297] The server stores the message data sent from the terminal in an analysis buffer.

[0298] Step 3:

[0299] The server begins parsing the message.

[0300] The server uses natural language processing (NLP) algorithms to tokenize and pattern match the message text.

[0301] Step 4:

[0302] An emotion engine analyzes the user's emotional state.

[0303] The emotion engine uses the user's facial expressions and behavioral data to assess their emotional state in real time.

[0304] Step 5:

[0305] The server assesses the likelihood of fraud.

[0306] The server determines the likelihood that the message content is fraudulent based on a registered list of fraudulent phrases and a reliability score.

[0307] Step 6:

[0308] Determines whether the server generates an alert.

[0309] If the possibility of fraud exceeds a threshold, the server generates an alert message based on the evaluation results of the emotion engine.

[0310] Step 7:

[0311] The device displays an alert to the user.

[0312] The device will display a pop-up warning message saying, "This email may be fraudulent. Please be careful."

[0313] Telephone fraud response features

[0314] Step 1:

[0315] Your device receives a call from a blocked or unknown number.

[0316] The device retrieves the caller information and matches it with an existing contact database.

[0317] Step 2:

[0318] The device will start an automatic voice response.

[0319] The terminal will begin the process of playing a pre-configured auto-answer message.

[0320] Step 3:

[0321] AI analyzes the content of calls in real time.

[0322] AI uses voice recognition technology to convert calls into text data, which is then analyzed using NLP algorithms.

[0323] Step 4:

[0324] The emotion engine analyzes the user's emotional state in real time.

[0325] The emotion engine analyzes the user's tone of voice and vocabulary to assess emotional fluctuations.

[0326] Step 5:

[0327] The server assesses the likelihood of fraud.

[0328] Based on the analysis results, the server scores the likelihood that the call content is fraudulent.

[0329] Step 6:

[0330] The server asks the user to confirm the call.

[0331] If the server determines that there is a high possibility of fraud, it generates an alert of appropriate strength based on the evaluation results of the emotion engine and sends it to the device.

[0332] Step 7:

[0333] The terminal displays a call confirmation to the user.

[0334] The device will display a message saying, "You have received an anonymous call. Please contact your family or the police to confirm the details."

[0335] Fraudulent wording learning function

[0336] Step 1:

[0337] The server collects historical fraud data.

[0338] The server stores past fraudulent emails and phone call details in a database.

[0339] Step 2:

[0340] The server learns the fraudulent language.

[0341] The server uses machine learning algorithms to analyze and extract common fraud patterns from the collected data.

[0342] Step 3:

[0343] The server updates the model.

[0344] The server adds and updates the learning model with newly discovered fraud techniques and patterns.

[0345] Step 4:

[0346] The emotion engine learns the user's emotion data.

[0347] The emotion engine learns the user's unique emotional patterns to more precisely assess the likelihood of fraud.

[0348] Step 5:

[0349] The server detects new fraud methods in real time.

[0350] The server applies the updated model to detect new fraud techniques from the latest data.

[0351] Step 6:

[0352] The terminal alerts the user based on new fraud patterns.

[0353] The device will display a notification saying, "This email may be a new scam. Do not click on any links."

[0354] The above are the processing steps in a specific embodiment of the present invention.

[0355] Example 2

[0356] 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."

[0357] In recent years, fraudulent activities via email and social networking services have been increasing, and many users have fallen victim to them. Fraudulent activities using anonymous and unknown phone numbers are also rampant. As fraudsters become more sophisticated, it is becoming increasingly difficult for users to distinguish fraudulent messages and calls, making it difficult to prevent damage before it occurs. In order to reduce the damage caused by fraud, a system that can detect possible fraud in real time and warn users appropriately is needed.

[0358] The specific processing by the specific processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means.

[0359] In this invention, the server includes means for receiving email messages and messages from social networking services, means for analyzing the received messages to detect potentially fraudulent wording and unreliable URLs, means for displaying an alert when there is a possibility of fraud, means for responding to calls from anonymous or unknown phone numbers with an automated voice and analyzing the content of the call in real time, means for learning from past fraud cases and updating the model to deal with new fraud methods, means for analyzing the user's emotional state and adjusting the strength and timing of the alert, and means for generating a warning message for the user based on the results of the analysis of the call content and messages. This makes it possible to detect fraudulent acts in real time and provide an optimal warning according to the user's emotional state.

[0360] An "email message" is an electronic document sent or received over the Internet that may contain text, images, links, etc.

[0361] "Social networking service" refers to a platform that enables users to interact and share information with other users on the Internet.

[0362] "Means of receiving" refers to the technical methods and equipment used to receive emails and messages into the user's terminal.

[0363] "Means for analyzing" refers to technical methods or devices that process the content of received messages or calls and analyze their content.

[0364] "Potentially fraudulent language" refers to words or phrases used with the intent to fraudulently obtain money or personal information.

[0365] "Untrusted URLs" refer to web links that are likely to be fraudulent or phishing.

[0366] "Means for displaying an alert" refers to a display device or method for issuing a warning or alert to the user.

[0367] "Hidden or unknown phone number" refers to a phone number where the caller's phone number is not displayed or is not registered by the user.

[0368] "Automated voice response means" means any technology or method that uses pre-recorded messages to answer telephone calls.

[0369] "Means of real-time analysis" refers to technologies and devices that process data the moment it is generated and output the results immediately.

[0370] "Past fraud cases" refers to data that records fraud cases and methods that have occurred in the past.

[0371] "Means for updating the model" refers to methods for updating machine learning algorithms and databases with the latest information.

[0372] "User's emotional state" refers to the psychological state or mood exhibited by a user at a particular moment.

[0373] "Means for adjusting the intensity and timing of alerts" refers to technologies and methods for appropriately changing the content and timing of warnings based on the user's emotional state.

[0374] "Means for generating a warning message" refers to a technique or method for creating and displaying a message to alert the user.

[0375] The present invention relates to a system for reducing the risk of fraudulent activity via email, social networking service (SNS) messages, and telephone calls. This system is implemented by a program with multiple functions. Each function is described in detail below, along with specific examples.

[0376] Email and SNS fraud detection function

[0377] When a device receives a new email or social media message, the server retrieves the message content and begins analyzing it. The server uses natural language processing (NLP) algorithms to detect potentially fraudulent content and unreliable URLs. Specific software that can be used is the popular NLP library "spaCy" or "NLTK." If a potential fraud is detected, an alert message is displayed on the device. An emotion engine is used to analyze the user's emotional state and adjust the intensity and content of the alert.

[0378] Examples:

[0379] 1. Your device will receive an email asking you to enter your bank account information.

[0380] 2. The server analyzes the text and detects possible fraud.

[0381] 3. The emotion engine analyzes the user's emotional state and sets the appropriate alert intensity.

[0382] 4. The device will pop up a warning message saying, "This email may be fraudulent. Please be careful."

[0383] Telephone fraud response features

[0384] When the device receives a call from a blocked or unknown phone number, it initiates an automated voice response. AI analyzes the call content in real time, and if there is a possibility of fraud, the server notifies the user of the call content and requests confirmation of the call. During this process, an emotion engine analyzes the user's emotional state in real time and adjusts the strength and timing of the warning based on emotional changes. The voice recognition software "Google Speech-to-Text API" and "Amazon Transcribe" can be used to analyze the call content.

[0385] Examples:

[0386] 1. The device receives an anonymous call.

[0387] 2. The device will respond with an automated voice saying, "This call is being recorded for security reasons. Please wait as this may be a scam."

[0388] 3. The AI ​​analyzes the content of the call and detects the phrase "I urgently need money."

[0389] 4. The emotion engine analyzes the user's emotional state during the call, and if the user feels anxious, the server generates an alert with appropriate intensity.

[0390] 5. The device will display the message, "You have received an anonymous call. Please contact your family or the police to confirm the details."

[0391] Fraudulent wording learning function

[0392] The server collects data on past fraud cases and uses machine learning algorithms to learn fraudulent phrases. This allows the model to be updated to deal with new fraud methods. Technologies used include machine learning libraries such as "scikit-learn" and "TensorFlow." This function makes it possible to detect unknown fraud methods early and take preventative measures. The emotion engine also learns from users' emotional data, understanding their unique emotional patterns to more precisely determine the likelihood of fraud.

[0393] Examples:

[0394] 1. The server analyzes past fraudulent emails and learns the common phrase "Please transfer the money quickly."

[0395] 2. The server adds the new fraud technique to the learning model.

[0396] 3. The server detects a new fraudulent email that says, "Click this link to receive a special discount."

[0397] 4. The emotion engine uses user emotional data to more precisely assess the likelihood of fraud.

[0398] 5. Your device will display a notification saying, "This email may be a new scam. Do not click on any links."

[0399] Example prompts to be input to the generative AI model

[0400] Detect fraudulent emails

[0401] "Assess the fraud risk of anonymous calls"

[0402] "Analyze the user's emotional state using an emotion engine."

[0403] The above is a specific embodiment of the system according to the present invention. The system of the present invention can detect fraudulent acts via email, social media, and telephone, and take appropriate action according to the user's emotional state, thereby preventing fraud damage before it occurs. This makes it possible to effectively reduce fraud damage.

[0404] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0405] Email and SNS fraud detection function

[0406] Processing Steps:

[0407] Step 1: Receive an email or social media message

[0408] When a device receives a new email or SNS message, it temporarily stores the contents in memory. This process requires no user input; the new message is the input data. The output is the message content itself.

[0409] Step 2: Parsing the message content

[0410] The server retrieves the received message content and analyzes it using a natural language processing (NLP) algorithm. Specifically, the message text is the input data, and the NLP algorithm detects potentially fraudulent phrases and unreliable URLs and assigns a score. The output is a score indicating the likelihood of fraud and the analysis results.

[0411] Step 3: Assess the likelihood of fraud

[0412] The server evaluates the fraud probability score based on the analysis results. If the score exceeds a certain threshold, it is determined that there is a high probability of fraud. The input is the analysis result from the previous step, and the output is the fraud risk assessment result.

[0413] Step 4: Alert the user

[0414] The terminal receives the fraud risk assessment results, and the emotion engine analyzes the user's emotional state and adjusts the strength and content of the alert. The input is the fraud risk assessment results and the user's emotional data, and the output is a warning display as a pop-up message. Specifically, the warning message is displayed in a pop-up window.

[0415] Telephone fraud response features

[0416] Processing Steps:

[0417] Step 1: Call from a blocked or unknown number

[0418] The terminal receives a call from a blocked or unknown number. The input is the incoming call signal and the output is the initiation of the call connection.

[0419] Step 2: Automated voice response

[0420] The terminal plays an automated voice message saying, "This call is being recorded for security reasons. Please wait as this may be a scam." The input is the start signal for the call connection, and the output is the automated voice message.

[0421] Step 3: Real-time analysis of call content

[0422] The AI ​​on the server analyzes the content of the call in real time and detects potentially fraudulent phrases. The input is the content of the call, and the output is the fraud risk assessment result. Specifically, the voice data is converted into text and analyzed.

[0423] Step 4: Emotional state analysis and alert generation

[0424] The emotion engine analyzes the user's emotional state during a call in real time and generates an alert based on the fraud risk assessment results. The input is the call content and emotional data, and the output is an alert message. Specifically, if the user feels anxious, a warning message will be displayed on the device.

[0425] Step 5: Notify users

[0426] The device displays the message "You have received an anonymous call. Please contact your family or the police to confirm the details." The input is the fraud risk assessment result, and the output is the display of a warning message.

[0427] Fraudulent wording learning function

[0428] Processing Steps:

[0429] Step 1: Collect data on past fraud cases

[0430] The server collects past fraudulent email and message cases from a database. The input is past fraud case data, and the output is a training dataset.

[0431] Step 2: Train using machine learning algorithms

[0432] The server uses a machine learning algorithm to learn the fraudulent claims: the input is a training dataset and the output is a trained model.

[0433] Step 3: Detect new fraud methods

[0434] The server analyzes new fraudulent messages in real time and determines the likelihood of fraud based on a learning model. The input is new message data, and the output is a fraud risk assessment result and a warning message.

[0435] Step 4: Learning emotion data

[0436] The emotion engine learns from the user's emotion data, understands the user's unique emotional patterns, and more precisely assesses fraud risk. The input is the user's emotion data, and the output is an updated emotion model.

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

[0438] Detect fraudulent emails

[0439] "Assess the fraud risk of anonymous calls"

[0440] "Analyze the user's emotional state using an emotion engine."

[0441] (Application example 2)

[0442] 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."

[0443] In modern society, frauds that exploit email and social networking services are on the rise. Furthermore, there are a wide variety of frauds that take place over the phone, resulting in an increasing number of victims. These frauds pose a significant risk, especially for those with low information literacy, such as the elderly. Furthermore, systems that issue uniform warnings without considering the user's emotional state make it difficult to effectively avoid fraud. Furthermore, as fraud methods evolve daily, conventional systems have the problem of being unable to respond quickly to new fraud methods.

[0444] The identification processing 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 receiving email messages and messages from social networking services; means for analyzing the received messages to detect potentially fraudulent wording or unreliable URLs; means for displaying an alert when there is a possibility of fraud; means for responding to calls from anonymous or unknown phone numbers with an automated voice and analyzing the content of the call in real time; means for learning from past fraud cases and updating a model to adapt to new fraud methods; means for analyzing the user's emotional state and adjusting the strength of the warning based on the emotional score; and means for providing a natural language processing algorithm for managing fraud risks in real time. This enables early detection of fraud risks and effective warnings based on the user's emotional state.

[0445] An "email message" is a communication containing information such as text, images, and links that is sent and received electronically over the Internet.

[0446] A "social networking service" is an online platform that enables people to share information, communicate, and interact over the Internet.

[0447] "Potential fraud" refers to situations in which a message or call received may have been sent with the intent to deceive the user.

[0448] An "untrusted URL" is an internet address that is likely to link to a fraudulent or malicious site.

[0449] The "means for displaying an alert" refers to a device or software that has the function of displaying a warning message or notification on a terminal to alert the user.

[0450] "Private or unknown phone number" refers to a phone number that does not display caller number information or is not registered in the user's contact list.

[0451] "Automatic voice response means" refers to the ability of the system to automatically respond to incoming calls with a pre-recorded voice message.

[0452] "Means for analyzing call content in real time" refers to a device or software that has the function of converting voice during a call into text and instantly analyzing the content.

[0453] "Means for learning from past fraud cases" refers to the ability to collect and analyze data on previous fraud cases and use that information to generate models that can respond to new fraud methods.

[0454] "Means for adjusting the strength of the warning based on the emotional score" refers to a function that quantifies the user's emotional state and adjusts the strength and display method of the warning message according to that score.

[0455] A "natural language processing algorithm" is a computational method for computers to understand, interpret, and generate human language.

[0456] An embodiment of a system that realizes this application example will be described below. This system has the function of receiving email messages and messages from social networking services (SNS) and analyzing their contents to detect fraud risks. It also has the function of responding to calls from blocked or unknown phone numbers with an automated voice and analyzing the contents of the call in real time. It also has the function of analyzing the user's emotional state and adjusting the strength of the warning based on the emotional score. The specific configuration and operation of this system will be described below.

[0457] The server has a means for receiving email messages and SNS messages. It retrieves messages using an email client or SNS API. These messages are then analyzed using a natural language processing algorithm (e.g., BERT) to detect potentially fraudulent content and unreliable URLs. If there is a possibility of fraud based on the detection results, a warning message is displayed on the device. At this time, the user's emotional state is analyzed and the strength of the warning is adjusted according to the emotional score. A sentiment analysis library such as VADER is used to analyze the emotional state.

[0458] When the device receives a call from a blocked or unknown number, it responds with an automated voice and analyzes the call in real time. It converts the call into text using technologies like Google Speech-to-Text and then analyzes it with natural language processing algorithms. If there is a possibility of fraud, the device notifies the user of the call and issues appropriate warnings if necessary.

[0459] The server collects and learns from past fraud cases and updates the machine learning model to adapt to new fraud techniques. This makes it possible to adapt to new fraud techniques. For example, the server can learn phrases such as "Please transfer the money quickly" and detect similar phrases in the future.

[0460] For example, if a user receives an email that asks them to enter their bank account information, the server analyzes the message to detect possible fraud. If emotion analysis reveals that the user is under stress, a stronger warning message will be displayed on the device: "This email may be fraudulent. Please be careful."

[0461] Furthermore, if an anonymous call is received and the content of the call is detected as "Please transfer 1 million yen immediately," the user will be shown a message saying, "You have received an anonymous call. Please contact your family or the police to confirm the content."

[0462] An example of a prompt sentence might be given to the generative AI model: "Analyze new emails and display a warning if there is any potentially fraudulent content. Vary the strength of the warning depending on the user's emotional state."

[0463] In this way, the system enables early detection of fraud risks and effective warnings based on the user's emotional state.

[0464] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[0465] Step 1: Receiving email and SNS messages

[0466] The server receives user email and SNS messages through email clients and SNS APIs. The input is new emails and SNS messages, and the output is the body of the received messages. Specifically, the server periodically accesses the email server or SNS platform to retrieve unread messages.

[0467] Step 2: Parsing the message content

[0468] The server analyzes the received message content using a natural language processing (NLP) algorithm (e.g., BERT). The input is the message body, and the output is a judgment result (e.g., a fraud score) on whether the message is likely to be fraudulent. Specifically, the message body is tokenized and input into an NLP model to evaluate the likelihood of fraud.

[0469] Step 3: Sentiment Analysis

[0470] The server analyzes the user's emotional state using an emotion analysis library such as VADER. The input is the user's message content and voice data, and the output is an emotion score. Specifically, it extracts emotional features from the message and voice data, inputs them into an emotion analysis model, and obtains an emotion score.

[0471] Step 4: Viewing warnings

[0472] If a potential fraud is detected, the device displays a warning message to the user. The strength of this warning is adjusted based on the emotion score. The input is the fraud score and the emotion score, and the output is the warning message. Specifically, the device compares the fraud score with the emotion score, generates an appropriate warning message, and displays it on the device's display.

[0473] Step 5: Auto-answering calls

[0474] When the device receives a call from a blocked or unknown phone number, it responds with an automated voice message. The input is the incoming call data, and the output is an automated voice response. Specifically, it plays a pre-recorded voice message and records the call in real time.

[0475] Step 6: Analyzing the call

[0476] The server converts the recorded conversations into text in real time and analyzes them using an NLP algorithm. The input is the call audio data, and the output is a judgment result on whether there is a possibility of fraud. Specifically, the audio data is converted into text using Google Speech-to-Text, and the text is then input into the NLP model for analysis.

[0477] Step 7: View call alerts

[0478] If a possible fraud is detected, the terminal displays a warning message about the call content to the user. The input is the analysis result, and the output is the call warning message. Specifically, the warning message is generated based on the analysis result and displayed on the terminal.

[0479] Step 8: Learn and update your scam language

[0480] The server periodically updates the machine learning model using past fraud cases. The input is past fraud data, and the output is an updated learning model. Specifically, the server uses collected fraud data to retrain the model so that it can adapt to new fraud patterns.

[0481] The above are the specific processing steps of the system that realizes the application example. At each step, appropriate data processing and calculations are performed based on the input data, and warnings and notifications are generated for the user.

[0482] 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.

[0483] 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.

[0484] 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.

[0485] [Second embodiment]

[0486] FIG. 3 shows an example of the configuration of a data processing system 210 according to the second embodiment.

[0487] 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.

[0488] 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).

[0489] 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.

[0490] 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.

[0491] 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).

[0492] 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.

[0493] 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.

[0494] 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.

[0495] 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.

[0496] 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.

[0497] 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."

[0498] The present invention relates to a system for reducing the risk of special fraud occurring through email, social networking services, and telephone. This system is realized by a program having the following functions.

[0499] Email and SNS fraud detection function

[0500] When a device receives a new email or message from a social networking service, the server analyzes the message's content. Specifically, the server uses natural language processing algorithms to analyze the message text and detect potentially fraudulent phrases and unreliable URLs. If the server determines that a message is potentially fraudulent, it displays an alert message on the device to warn the user.

[0501] Examples:

[0502] 1. Your device will receive an email asking you to enter your bank account information.

[0503] 2. The server analyzes the text and detects possible fraud.

[0504] 3. The device notifies the user, "This email may be fraudulent. Please be careful."

[0505] Telephone fraud response features

[0506] When the device receives a call from a blocked or unknown number, the device will initiate an automated voice response. AI will analyze the call content in real time, and if there is a possibility of fraud, the server will notify the user of the call content and request confirmation of the call.

[0507] Examples:

[0508] 1. The device receives an anonymous call.

[0509] 2. The device will respond with an automated voice saying, "This call is being recorded for security reasons. Please wait as this may be a scam."

[0510] 3. The AI ​​analyzes the content of the call and detects the phrase "I urgently need money."

[0511] 4. The server detects possible fraud and

[0512] 5. The device notifies the user, "You have received an anonymous call. Please contact your family or the police to confirm the details."

[0513] Fraudulent wording learning function

[0514] The server collects data on past fraud cases and uses machine learning algorithms to learn fraudulent phrases. This allows the model to be updated to handle new fraud methods and enable early detection of unknown fraud methods. If a new fraud method is detected, the device will display a warning to the user based on that information.

[0515] Examples:

[0516] 1. The server analyzes past fraudulent emails and learns the common phrase "Please transfer the money quickly."

[0517] 2. The server adds the new fraud technique to the learning model.

[0518] 3. The server detects a new fraudulent email that says, "Click this link to receive a special discount."

[0519] 4. The device notifies the user, "This email may be a new scam. Do not click on the link."

[0520] As described above, the system according to the present invention can detect fraudulent acts via email, social media, and telephone, and can prevent fraud damage by issuing a warning to the user in advance, thereby effectively reducing fraud damage.

[0521] The processing flow will be explained below.

[0522] Email and SNS fraud detection function

[0523] Step 1:

[0524] Your device receives a new email or SNS message.

[0525] The device will notify you when a new message arrives in your inbox.

[0526] Step 2:

[0527] Gets the contents of the message received by the server.

[0528] The server stores the message data sent from the terminal in an analysis buffer.

[0529] Step 3:

[0530] The server begins parsing the message.

[0531] The server uses natural language processing (NLP) algorithms to tokenize and pattern match the message text.

[0532] Step 4:

[0533] The server assesses the likelihood of fraud.

[0534] The server determines the likelihood that the message content is fraudulent based on a registered list of fraudulent phrases and a reliability score.

[0535] Step 5:

[0536] Determines whether the server generates an alert.

[0537] The server decides to generate an alert message if the likelihood of fraud exceeds a threshold.

[0538] Step 6:

[0539] The device displays an alert to the user.

[0540] The device will display a pop-up warning message saying, "This email may be fraudulent. Please be careful."

[0541] Telephone fraud response features

[0542] Step 1:

[0543] Your device receives a call from a blocked or unknown number.

[0544] The device retrieves the caller information and matches it with an existing contact database.

[0545] Step 2:

[0546] The device will start an automatic voice response.

[0547] The terminal will begin the process of playing a pre-configured auto-answer message.

[0548] Step 3:

[0549] AI analyzes the content of calls in real time.

[0550] AI uses voice recognition technology to convert calls into text data, which is then analyzed using NLP algorithms.

[0551] Step 4:

[0552] The server assesses the likelihood of fraud.

[0553] Based on the analysis results, the server scores the likelihood that the call content is fraudulent.

[0554] Step 5:

[0555] The server asks the user to confirm the call.

[0556] If the server determines that there is a high possibility of fraud, it sends a warning message to the user's terminal.

[0557] Step 6:

[0558] The terminal displays a call confirmation to the user.

[0559] The device will display a message saying, "You have received an anonymous call. Please contact your family or the police to confirm the details."

[0560] Fraudulent wording learning function

[0561] Step 1:

[0562] The server collects historical fraud data.

[0563] The server stores past fraudulent emails and phone call details in a database.

[0564] Step 2:

[0565] The server learns the fraudulent language.

[0566] The server uses machine learning algorithms to analyze and extract common fraud patterns from the collected data.

[0567] Step 3:

[0568] The server updates the model.

[0569] The server adds and updates the learning model with newly discovered fraud techniques and patterns.

[0570] Step 4:

[0571] The server detects new fraud methods in real time.

[0572] The server applies the updated model to detect new fraud techniques from the latest data.

[0573] Step 5:

[0574] The terminal alerts the user based on new fraud patterns.

[0575] The device will display a notification saying, "This email may be a new scam. Do not click on any links."

[0576] The above are the processing steps in a specific embodiment of the present invention.

[0577] Example 1

[0578] 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."

[0579] Conventional systems have had difficulty effectively reducing the risk of fraud via email, social media, and telephone. In particular, they were unable to adapt to new fraud methods, increasing the likelihood of fraud victims occurring. Furthermore, users often felt uneasy because they were unable to analyze call content in real time to detect potential fraud.

[0580] 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.

[0581] In this invention, the server includes: means for receiving email messages and messages from social networking services; means for analyzing the received messages to detect potentially fraudulent wording and unreliable URLs; means for displaying an alert when there is a possibility of fraud; means for responding to calls from anonymous or unknown phone numbers with an automated voice and analyzing the content of the call in real time; means for learning from past fraud cases and updating the model to respond to new fraud techniques; and means for displaying a warning to the user when there is a possibility of fraud among the above means; means for analyzing the content of messages and calls in real time using a natural language processing algorithm to detect possible fraud; and means for detecting new fraud techniques using a generative AI model based on the analysis results. This makes it possible to detect fraudulent acts via email, social networking services, and telephone calls early and prevent fraud damage by issuing advance warnings to users.

[0582] "Email message" refers to the text information of an email sent or received over the Internet.

[0583] A "social networking service" is a platform that allows people to share information and communicate online.

[0584] "Means of receiving" refers to the technology and equipment used to receive email messages and messages from social networking services.

[0585] "Means for analysis" refers to the technology and equipment used to analyze the content of received messages and find specific patterns or characteristics.

[0586] "Potentially fraudulent language and unreliable URLs" refers to keywords associated with fraudulent activities and unreliable web links.

[0587] "Means for displaying an alert" refers to the technology and devices used to display a warning message to the user when a potential fraud is detected.

[0588] "Anonymous or unknown phone numbers" are phone numbers that cannot be identified by the caller and that have not been previously registered.

[0589] "Automatic voice response means" means technology and equipment that automatically plays a voice message in response to an incoming call without human intervention.

[0590] "Means for analyzing call content in real time" refers to technology and equipment that allows for immediate analysis of the content of a call while it is being made.

[0591] "Past fraud cases" refers to specific cases and data of fraudulent acts that have occurred in the past.

[0592] "Model updating methods" refers to the techniques and methods used to keep fraud detection algorithms up to date based on new information and data.

[0593] A "natural language processing algorithm" is an algorithm that allows a computer to analyze and understand human language.

[0594] A "generative AI model" is an artificial intelligence model that uses machine learning techniques to generate new patterns and knowledge from data.

[0595] "Means for displaying a warning to the user" refers to technologies and devices that provide visual or audio warnings to the user when possible fraud is detected.

[0596] The present invention relates to a system for reducing the risk of special frauds occurring through email, social networking services (SNS), and telephone. This system includes functions for receiving email and SNS messages, analyzing the received messages, determining the possibility of fraud, displaying alerts, responding to telephone fraud, and learning fraudulent phrases.

[0597] Email and SNS fraud detection function

[0598] The server receives emails and SNS messages using an internet-connected device (smartphone, PC, tablet, etc.). The received message is sent to the server via a secure communication protocol (e.g., HTTPS). The server analyzes the content of the message using a natural language processing algorithm (e.g., spaCy, NLTK) to detect specific keywords (e.g., 'bank account information', 'transfer') and unreliable URLs. If it determines that there is a possibility of fraud, an alert message is generated and sent to the device. The device notifies the user, "This email may be fraudulent. Please be careful."

[0599] Specific examples

[0600] When the terminal receives an email saying "Please enter your bank account information," it sends the received information to the server.

[0601] The server analyzes the email and detects possible fraud.

[0602] The device will notify the user, "This email may be fraudulent. Please be careful."

[0603] Prompt Sentence Examples

[0604] Here's a fraud warning message for when you receive an email that says "enter your bank account details."

[0605] Telephone fraud response features

[0606] When the device receives a call from an anonymous or unknown phone number, an automated voice response begins, saying, "This call is being recorded for safety reasons. Please wait as this may be a scam." The content of the call is analyzed in real time by AI, and the results are sent to the server. The server analyzes the call text, and if it detects phrases that are likely to be fraudulent (e.g., 'I urgently need money'), it sends a notification to the user requesting that they check the content of the call. The device notifies the user, "You have received an anonymous call. Please contact your family or the police to confirm the content."

[0607] Specific examples

[0608] When the device receives an anonymous call, an automated voice responds saying, "This call is being recorded for safety reasons. Please wait as this may be a scam."

[0609] AI analyzes the content of calls in real time and detects phrases such as "I urgently need money."

[0610] The server determines the possibility of fraud, and the device notifies the user, "You have received an anonymous call. Please contact your family or the police to confirm the details."

[0611] Prompt Sentence Examples

[0612] Display a warning message if an anonymous caller claims to be in urgent need of money.

[0613] Fraudulent wording learning function

[0614] The server collects data on past fraud cases and analyzes it using machine learning algorithms (e.g., TensorFlow, Scikit-learn). It learns characteristic patterns and keywords from the analyzed data and updates the model to respond to new fraud methods. The server detects new fraud methods and sends a notification to the device. The device displays a warning to the user saying, "This email may be a new scam. Do not click on the link."

[0615] Specific examples

[0616] The server analyzes past fraudulent emails and learns the phrase "Please transfer the money quickly."

[0617] The server adds new fraud techniques to the learning model.

[0618] The server now detects the phrase "Click this link to receive a special discount."

[0619] The device will notify the user, "This email may be a new scam. Do not click on the link."

[0620] Prompt Sentence Examples

[0621] A fraud warning message for those who receive emails that say, "Click this link to receive a special discount."

[0622] This system can detect fraudulent activities via email, social media, and telephone at an early stage and warn users in advance, preventing fraud damage before it occurs. This will significantly reduce the damage caused by fraud.

[0623] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0624] Email and SNS fraud detection function

[0625] Processing Steps

[0626] Step 1:

[0627] The device receives new emails or SNS messages and saves their contents to local storage. The input is the received message, and the output is the saved message data.

[0628] Step 2:

[0629] The terminal sends the received message to the server using a secure communication protocol (e.g. HTTPS). The input is the stored message data, and the output is the message data sent to the server.

[0630] Step 3:

[0631] The server analyzes the message data it receives. Specifically, it uses a natural language processing library (e.g., spaCy, NLTK) to tokenize the message, tag parts of speech, and analyze context. The input is the message data sent to the server, and the output is the analysis results.

[0632] Step 4:

[0633] The server determines the likelihood of fraud based on the analysis results. Criteria for determining fraud include specific keywords (e.g., 'bank account information', 'transfer') and untrustworthy URLs. The input is the analysis results, and the output is an assessment of the likelihood of fraud.

[0634] Step 5:

[0635] If the server detects a possibility of fraud, it generates an alert message and sends it to the terminal. The input is the evaluation result of the possibility of fraud, and the output is the generated alert message.

[0636] Step 6:

[0637] The terminal notifies the user of the alert message. The input is the generated alert message, and the output is the warning message displayed to the user.

[0638] Telephone fraud response features

[0639] Processing Steps

[0640] Step 1:

[0641] The terminal receives a call from a blocked or unknown phone number. The input is the call from the blocked or unknown phone number, and the output is the acceptance of the call.

[0642] Step 2:

[0643] The terminal starts an automated voice response, saying, "This call is being recorded for security reasons. Please wait as this may be a scam." The input is the acceptance of the call, and the output is the automated voice response.

[0644] Step 3:

[0645] AI analyzes the contents of calls in real time. Specifically, it converts the contents of calls into text using the Google Speech-to-Text API and performs keyword analysis. The input is the contents of the calls, and the output is the analysis results.

[0646] Step 4:

[0647] The server determines the likelihood of fraud based on the analysis results. Fraud criteria include specific phrases (e.g., 'I urgently need money'). The input is the analysis results, and the output is an assessment of the likelihood of fraud.

[0648] Step 5:

[0649] If the server detects the possibility of fraud, it generates a notification requesting the user to confirm the contents of the call and sends it to the terminal. The input is the result of the evaluation of the possibility of fraud, and the output is the generated notification.

[0650] Step 6:

[0651] The device sends a notification to the user, warning them that "An anonymous call has been made. Please contact your family or the police to confirm the details." The input is the generated notification, and the output is the warning message that is displayed to the user.

[0652] Fraudulent wording learning function

[0653] Processing Steps

[0654] Step 1:

[0655] The server collects past fraudulent email and call data from the Internet and internal databases. The input is past fraud case data, and the output is the collected data.

[0656] Step 2:

[0657] The server uses machine learning algorithms (e.g., TensorFlow, Scikit-learn) to analyze the collected data and learn specific patterns or keywords. The input is the collected data, and the output is an updated learning model.

[0658] Step 3:

[0659] The server adds new fraud techniques to the learning model. The input is specific patterns or keywords, and the output is an updated learning model.

[0660] Step 4:

[0661] The server analyzes newly received emails and messages to detect new fraud methods. The input is the new email or message, and the output is the evaluation result of the likelihood of fraud.

[0662] Step 5:

[0663] If the device detects a possible scam, it notifies the user, "This email contains a new potential scam. Do not click on the link." The input is the result of the evaluation of the possible scam, and the output is a warning message that is displayed to the user.

[0664] (Application example 1)

[0665] 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."

[0666] In modern society, fraudulent activities via email, social networking services (SNS), and telephone are on the rise, and many people are falling victim to these scams. Elderly people and those unfamiliar with the Internet have particular difficulty distinguishing between fraudulent messages and phone calls, making it easier for the damage to spread. There is a need for technology that can effectively detect such fraudulent activities and issue early warnings to users.

[0667] 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.

[0668] In this invention, the server includes means for receiving email messages and messages from social networking services, means for analyzing the received messages to detect potentially fraudulent wording and unreliable URLs, means for displaying an alert when there is a possibility of fraud, means for responding to calls from withheld or unknown phone numbers with an automated voice and analyzing the content of the call in real time, means for learning from past fraud cases and updating the model to deal with new fraud methods, and means for visually warning users of the risk of fraud using a display device worn by the user. This makes it easier for users to recognize the risk of fraud in real time and prevents them from becoming victims of fraud.

[0669] "Mail message" refers to information, including text and attachments, sent and received via email.

[0670] A "social networking service" is an online platform that enables people to share information and deepen their interactions via the Internet.

[0671] A "server" is a computer system that processes and provides data over a network.

[0672] "Means for receiving messages" refers to technology that has the function of receiving emails and SNS messages on a user's device.

[0673] "Means for analyzing" refers to technology for analyzing the content of received messages to assess fraud risk.

[0674] "Wording" refers to a sentence or phrase, and refers to specific text that may be fraudulent.

[0675] "URL" stands for Uniform Resource Locator and is a description that indicates the address to a web page or online resource.

[0676] "Means for displaying an alert" refers to technology that displays a warning or alert message to the user.

[0677] "Calls from blocked or unknown numbers" refers to calls from phones that do not display or are not registered and cannot be identified.

[0678] "Means for responding with automated voice" refers to technology in which a system automatically generates and responds with a voice message.

[0679] "Means for analyzing call content in real time" refers to technology that instantly converts voice data during a call into text, analyzes the content, and assesses the risk of fraud.

[0680] "Past fraud cases" refers to data on fraud methods and cases that have been reported to date.

[0681] "Means of learning and updating models to adapt to new fraud techniques" refers to a technique that uses machine learning algorithms to learn new fraud patterns and update existing detection models.

[0682] A "display device" refers to hardware for visually presenting information, such as smart contact lenses or smart glasses.

[0683] "Means for visually warning of fraud risk" refers to technology that uses a visual display device worn by the user to display a warning about fraud risk.

[0684] The present invention relates to a system for reducing the risk of special frauds occurring through email, social networking services (SNS), and telephone. This system is realized by having the following main functions.

[0685] 1. Email and SNS fraud detection function

[0686] When a device receives a new email or message from a social networking service, the server analyzes the message's content. Specifically, the server uses a natural language processing library (e.g., Spacy) and a machine learning model (e.g., a model saved by scikit-learn) to analyze the message text and detect potentially fraudulent phrases or unreliable URLs. If any are detected, a warning message is displayed on the device in real time to alert the user.

[0687] 2. Telephone fraud response function

[0688] When the device receives a call from a blocked or unknown number, the device will initiate an automated voice response. AI will analyze the call content in real time, and if there is a possibility of fraud, the server will notify the user of the call content and request confirmation of the call.

[0689] 3. Fraudulent wording learning function

[0690] The server collects data on past fraud cases and uses machine learning algorithms to learn fraudulent phrases. This allows the model to be updated to handle new fraud methods and enable early detection of unknown fraud methods. If a new fraud method is detected, the device will display a warning to the user based on that information.

[0691] 4. Warning display function on smart contact lenses

[0692] The system also includes a function that visually warns users of fraud risks using display devices such as smart contact lenses or smart glasses. If the server detects a risk of fraud while the user is checking email or social media messages, a visual warning message saying "Possible fraud" will be displayed on the display device.

[0693] Hardware / software configuration used

[0694] Server: Analyzes messages, runs AI models, and updates learning models.

[0695] Device: Receives emails and SNS messages, records phone calls, and provides automated voice responses.

[0696] Natural language processing libraries such as Spacy can be used to analyze message content.

[0697] Machine learning models: Assess fraud risk using models saved in Scikit-learn.

[0698] Display devices: Use smart contact lenses or smart glasses to display warning messages.

[0699] Specific examples

[0700] For example, if a user receives an email requesting "Please enter your bank account information," the server analyzes the text and detects the possibility of fraud. The device (such as a smartphone) then displays a warning message saying, "This email may be fraudulent. Please be careful." At the same time, if the user is wearing smart contact lenses, the lenses will also display a visual warning saying, "This email may be fraudulent."

[0701] Prompt Sentence Examples

[0702] An example prompt for new messages would look something like this:

[0703] "A new message has arrived. Please analyze its contents."

[0704] As a result, this system can effectively detect fraudulent activities via email, social media, and telephone, and can prevent fraud damage by issuing advance warnings to users, thereby effectively reducing fraud damage.

[0705] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0706] Step 1:

[0707] The device receives a new email or message from a social networking service. The received message data is provided as input. The device sends this message to the server.

[0708] Step 2:

[0709] The server analyzes the content of the received message. The input is the text data of the received message, and the output is the analysis result. The server uses a natural language processing library (e.g., Spacy) to tokenize the message text and detect potentially fraudulent phrases and unreliable URLs.

[0710] Step 3:

[0711] The server determines the likelihood of fraud based on the analysis results. The input is the message analysis results, and the output is the fraud risk assessment result. The fraud risk is assessed using a machine learning model trained on past fraud cases.

[0712] Step 4:

[0713] If there is a possibility of fraud, the server sends a warning to the terminal and display device. The input is the fraud risk assessment result (if the fraud risk is high), and the output is a warning message to be displayed to the user. The server generates a warning message such as "Possible fraud" and sends it to the terminal and display device such as a smart contact lens.

[0714] Step 5:

[0715] When a user receives a call from a blocked or unknown phone number, the device starts an automated voice response. The input is the incoming call information, and the output is an automated voice message. The device generates an automated voice message informing the user, "This call is being recorded for safety reasons. Please wait as this may be a scam."

[0716] Step 6:

[0717] The server analyzes the call content in real time. The input is the voice data of the call, and the output is the text conversion and analysis results of the call content. The server uses AI to convert the call content into text and analyze it.

[0718] Step 7:

[0719] The server evaluates the possibility of fraud based on the analysis results of the call content. The input is the text data of the call content and the analysis results, and the output is the fraud risk assessment result.

[0720] Step 8:

[0721] If there is a possibility of fraud, the server sends a warning to the terminal and display device and requests the user to confirm the call. The input is the fraud risk assessment result (if the fraud risk is high), and the output is a warning message to be displayed to the user and a request to confirm the content of the call. The server generates a warning message saying "There has been an anonymous call. Please consult with your family or the police to confirm the content," and sends it to the terminal and display device.

[0722] Step 9:

[0723] The server collects data on past fraud cases and uses a machine learning algorithm to learn fraudulent phrases. The input is data on past fraud cases, and the output is an updated fraud detection model. If a new fraud technique is detected, it is added to the learning model and reflected in the next analysis.

[0724] Through the above processing steps, this system can effectively detect fraudulent activities via email, social media, and telephone, and can prevent fraudulent acts by issuing advance warnings to users.

[0725] 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.

[0726] The present invention combines an emotion engine with a system for reducing the risk of special frauds occurring through email, social networking services, and telephone calls, enabling appropriate responses to be taken according to the user's emotional state. This system is realized by a program with the following functions:

[0727] Email and SNS fraud detection function

[0728] When a device receives a new email or message from a social networking service, the server retrieves and analyzes the message content. Specifically, the server uses a natural language processing (NLP) algorithm to analyze the message text and detect potentially fraudulent language or unreliable URLs. If the possibility of fraud is assessed, an alert message is displayed on the device to warn the user. At this time, the emotion engine recognizes the user's emotional state, and if the user is in a state of stress, a stronger alert is displayed.

[0729] Examples:

[0730] 1. Your device will receive an email asking you to enter your bank account information.

[0731] 2. The server analyzes the text and detects possible fraud.

[0732] 3. The emotion engine analyzes the user's emotional state and sets the appropriate alert intensity.

[0733] 4. The device will pop up a warning message saying, "This email may be fraudulent. Please be careful."

[0734] Telephone fraud response features

[0735] When a device receives a call from a blocked or unknown number, it will initiate an automated voice response. AI analyzes the call content in real time, and if there is a possibility of fraud, the server will notify the user of the call content and request confirmation. During this process, an emotion engine analyzes the user's emotional state in real time and adjusts the intensity and timing of the warning based on emotional changes.

[0736] Examples:

[0737] 1. The device receives an anonymous call.

[0738] 2. The device will respond with an automated voice saying, "This call is being recorded for security reasons. Please wait as this may be a scam."

[0739] 3. The AI ​​analyzes the content of the call and detects the phrase "I urgently need money."

[0740] 4. The emotion engine analyzes the user's emotional state during the call, and if the user feels anxious, the server generates an alert with appropriate intensity.

[0741] 5. The device will display the message, "You have received an anonymous call. Please contact your family or the police to confirm the details."

[0742] Fraudulent wording learning function

[0743] The server collects data on past fraud cases and uses machine learning algorithms to learn fraudulent phrases. This allows the model to be updated to handle new fraud methods, enabling early detection of unknown fraud methods. The emotion engine also learns from the user's emotional data, allowing it to understand the user's unique emotional patterns and more precisely determine the likelihood of fraud. If a new fraud method is detected, the device will display a warning to the user based on that information.

[0744] Examples:

[0745] 1. The server analyzes past fraudulent emails and learns the common phrase "Please transfer the money quickly."

[0746] 2. The server adds the new fraud technique to the learning model.

[0747] 3. The server detects a new fraudulent email that says, "Click this link to receive a special discount."

[0748] 4. The emotion engine uses user emotional data to more precisely assess the likelihood of fraud.

[0749] 5. Your device will display a notification saying, "This email may be a new scam. Do not click on any links."

[0750] As described above, the system according to the present invention can detect fraudulent acts via email, social media, and telephone, and take appropriate measures according to the user's emotional state to prevent fraud damage before it occurs, thereby effectively reducing fraud damage.

[0751] The processing flow will be explained below.

[0752] Email and SNS fraud detection function

[0753] Step 1:

[0754] Your device receives a new email or SNS message.

[0755] The device will notify you when a new message arrives in your inbox.

[0756] Step 2:

[0757] Gets the contents of the message received by the server.

[0758] The server stores the message data sent from the terminal in an analysis buffer.

[0759] Step 3:

[0760] The server begins parsing the message.

[0761] The server uses natural language processing (NLP) algorithms to tokenize and pattern match the message text.

[0762] Step 4:

[0763] An emotion engine analyzes the user's emotional state.

[0764] The emotion engine uses the user's facial expressions and behavioral data to assess their emotional state in real time.

[0765] Step 5:

[0766] The server assesses the likelihood of fraud.

[0767] The server determines the likelihood that the message content is fraudulent based on a registered list of fraudulent phrases and a reliability score.

[0768] Step 6:

[0769] Determines whether the server generates an alert.

[0770] If the possibility of fraud exceeds a threshold, the server generates an alert message based on the evaluation results of the emotion engine.

[0771] Step 7:

[0772] The device displays an alert to the user.

[0773] The device will display a pop-up warning message saying, "This email may be fraudulent. Please be careful."

[0774] Telephone fraud response features

[0775] Step 1:

[0776] Your device receives a call from a blocked or unknown number.

[0777] The device retrieves the caller information and matches it with an existing contact database.

[0778] Step 2:

[0779] The device will start an automatic voice response.

[0780] The terminal will begin the process of playing a pre-configured auto-answer message.

[0781] Step 3:

[0782] AI analyzes the content of calls in real time.

[0783] AI uses voice recognition technology to convert calls into text data, which is then analyzed using NLP algorithms.

[0784] Step 4:

[0785] The emotion engine analyzes the user's emotional state in real time.

[0786] The emotion engine analyzes the user's tone of voice and vocabulary to assess emotional fluctuations.

[0787] Step 5:

[0788] The server assesses the likelihood of fraud.

[0789] Based on the analysis results, the server scores the likelihood that the call content is fraudulent.

[0790] Step 6:

[0791] The server asks the user to confirm the call.

[0792] If the server determines that there is a high possibility of fraud, it generates an alert of appropriate strength based on the evaluation results of the emotion engine and sends it to the device.

[0793] Step 7:

[0794] The terminal displays a call confirmation to the user.

[0795] The device will display a message saying, "You have received an anonymous call. Please contact your family or the police to confirm the details."

[0796] Fraudulent wording learning function

[0797] Step 1:

[0798] The server collects historical fraud data.

[0799] The server stores past fraudulent emails and phone call details in a database.

[0800] Step 2:

[0801] The server learns the fraudulent language.

[0802] The server uses machine learning algorithms to analyze and extract common fraud patterns from the collected data.

[0803] Step 3:

[0804] The server updates the model.

[0805] The server adds and updates the learning model with newly discovered fraud techniques and patterns.

[0806] Step 4:

[0807] The emotion engine learns the user's emotion data.

[0808] The emotion engine learns the user's unique emotional patterns to more precisely assess the likelihood of fraud.

[0809] Step 5:

[0810] The server detects new fraud methods in real time.

[0811] The server applies the updated model to detect new fraud techniques from the latest data.

[0812] Step 6:

[0813] The terminal alerts the user based on new fraud patterns.

[0814] The device will display a notification saying, "This email may be a new scam. Do not click on any links."

[0815] The above are the processing steps in a specific embodiment of the present invention.

[0816] Example 2

[0817] 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."

[0818] In recent years, fraudulent activities via email and social networking services have been increasing, and many users have fallen victim to them. Fraudulent activities using anonymous and unknown phone numbers are also rampant. As fraudsters become more sophisticated, it is becoming increasingly difficult for users to distinguish fraudulent messages and calls, making it difficult to prevent damage before it occurs. In order to reduce the damage caused by fraud, a system that can detect possible fraud in real time and warn users appropriately is needed.

[0819] The specific processing by the specific processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means.

[0820] In this invention, the server includes means for receiving email messages and messages from social networking services, means for analyzing the received messages to detect potentially fraudulent wording and unreliable URLs, means for displaying an alert when there is a possibility of fraud, means for responding to calls from anonymous or unknown phone numbers with an automated voice and analyzing the content of the call in real time, means for learning from past fraud cases and updating the model to deal with new fraud methods, means for analyzing the user's emotional state and adjusting the strength and timing of the alert, and means for generating a warning message for the user based on the results of the analysis of the call content and messages. This makes it possible to detect fraudulent acts in real time and provide an optimal warning according to the user's emotional state.

[0821] An "email message" is an electronic document sent or received over the Internet that may contain text, images, links, etc.

[0822] "Social networking service" refers to a platform that enables users to interact and share information with other users on the Internet.

[0823] "Means of receiving" refers to the technical methods and equipment used to receive emails and messages into the user's terminal.

[0824] "Means for analyzing" refers to technical methods or devices that process the content of received messages or calls and analyze their content.

[0825] "Potentially fraudulent language" refers to words or phrases used with the intent to fraudulently obtain money or personal information.

[0826] "Untrusted URLs" refer to web links that are likely to be fraudulent or phishing.

[0827] "Means for displaying an alert" refers to a display device or method for issuing a warning or alert to the user.

[0828] "Hidden or unknown phone number" refers to a phone number where the caller's phone number is not displayed or is not registered by the user.

[0829] "Automated voice response means" means any technology or method that uses pre-recorded messages to answer telephone calls.

[0830] "Means of real-time analysis" refers to technologies and devices that process data the moment it is generated and output the results immediately.

[0831] "Past fraud cases" refers to data that records fraud cases and methods that have occurred in the past.

[0832] "Means for updating the model" refers to methods for updating machine learning algorithms and databases with the latest information.

[0833] "User's emotional state" refers to the psychological state or mood exhibited by a user at a particular moment.

[0834] "Means for adjusting the intensity and timing of alerts" refers to technologies and methods for appropriately changing the content and timing of warnings based on the user's emotional state.

[0835] "Means for generating a warning message" refers to a technique or method for creating and displaying a message to alert the user.

[0836] The present invention relates to a system for reducing the risk of fraudulent activity via email, social networking service (SNS) messages, and telephone calls. This system is implemented by a program with multiple functions. Each function is described in detail below, along with specific examples.

[0837] Email and SNS fraud detection function

[0838] When a device receives a new email or social media message, the server retrieves the message content and begins analyzing it. The server uses natural language processing (NLP) algorithms to detect potentially fraudulent content and unreliable URLs. Specific software that can be used is the popular NLP library "spaCy" or "NLTK." If a potential fraud is detected, an alert message is displayed on the device. An emotion engine is used to analyze the user's emotional state and adjust the intensity and content of the alert.

[0839] Examples:

[0840] 1. Your device will receive an email asking you to enter your bank account information.

[0841] 2. The server analyzes the text and detects possible fraud.

[0842] 3. The emotion engine analyzes the user's emotional state and sets the appropriate alert intensity.

[0843] 4. The device will pop up a warning message saying, "This email may be fraudulent. Please be careful."

[0844] Telephone fraud response features

[0845] When the device receives a call from a blocked or unknown phone number, it initiates an automated voice response. AI analyzes the call content in real time, and if there is a possibility of fraud, the server notifies the user of the call content and requests confirmation of the call. During this process, an emotion engine analyzes the user's emotional state in real time and adjusts the strength and timing of the warning based on emotional changes. The voice recognition software "Google Speech-to-Text API" and "Amazon Transcribe" can be used to analyze the call content.

[0846] Examples:

[0847] 1. The device receives an anonymous call.

[0848] 2. The device will respond with an automated voice saying, "This call is being recorded for security reasons. Please wait as this may be a scam."

[0849] 3. The AI ​​analyzes the content of the call and detects the phrase "I urgently need money."

[0850] 4. The emotion engine analyzes the user's emotional state during the call, and if the user feels anxious, the server generates an alert with appropriate intensity.

[0851] 5. The device will display the message, "You have received an anonymous call. Please contact your family or the police to confirm the details."

[0852] Fraudulent wording learning function

[0853] The server collects data on past fraud cases and uses machine learning algorithms to learn fraudulent phrases. This allows the model to be updated to deal with new fraud methods. Technologies used include machine learning libraries such as "scikit-learn" and "TensorFlow." This function makes it possible to detect unknown fraud methods early and take preventative measures. The emotion engine also learns from users' emotional data, understanding their unique emotional patterns to more precisely determine the likelihood of fraud.

[0854] Examples:

[0855] 1. The server analyzes past fraudulent emails and learns the common phrase "Please transfer the money quickly."

[0856] 2. The server adds the new fraud technique to the learning model.

[0857] 3. The server detects a new fraudulent email that says, "Click this link to receive a special discount."

[0858] 4. The emotion engine uses user emotional data to more precisely assess the likelihood of fraud.

[0859] 5. Your device will display a notification saying, "This email may be a new scam. Do not click on any links."

[0860] Example prompts to be input to the generative AI model

[0861] Detect fraudulent emails

[0862] "Assess the fraud risk of anonymous calls"

[0863] "Analyze the user's emotional state using an emotion engine."

[0864] The above is a specific embodiment of the system according to the present invention. The system of the present invention can detect fraudulent acts via email, social media, and telephone, and take appropriate action according to the user's emotional state, thereby preventing fraud damage before it occurs. This makes it possible to effectively reduce fraud damage.

[0865] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0866] Email and SNS fraud detection function

[0867] Processing Steps:

[0868] Step 1: Receive an email or social media message

[0869] When a device receives a new email or SNS message, it temporarily stores the contents in memory. This process requires no user input; the new message is the input data. The output is the message content itself.

[0870] Step 2: Parsing the message content

[0871] The server retrieves the received message content and analyzes it using a natural language processing (NLP) algorithm. Specifically, the message text is the input data, and the NLP algorithm detects potentially fraudulent phrases and unreliable URLs and assigns a score. The output is a score indicating the likelihood of fraud and the analysis results.

[0872] Step 3: Assess the likelihood of fraud

[0873] The server evaluates the fraud probability score based on the analysis results. If the score exceeds a certain threshold, it is determined that there is a high probability of fraud. The input is the analysis result from the previous step, and the output is the fraud risk assessment result.

[0874] Step 4: Alert the user

[0875] The terminal receives the fraud risk assessment results, and the emotion engine analyzes the user's emotional state and adjusts the strength and content of the alert. The input is the fraud risk assessment results and the user's emotional data, and the output is a warning display as a pop-up message. Specifically, the warning message is displayed in a pop-up window.

[0876] Telephone fraud response features

[0877] Processing Steps:

[0878] Step 1: Call from a blocked or unknown number

[0879] The terminal receives a call from a blocked or unknown number. The input is the incoming call signal and the output is the initiation of the call connection.

[0880] Step 2: Automated voice response

[0881] The terminal plays an automated voice message saying, "This call is being recorded for security reasons. Please wait as this may be a scam." The input is the start signal for the call connection, and the output is the automated voice message.

[0882] Step 3: Real-time analysis of call content

[0883] The AI ​​on the server analyzes the content of the call in real time and detects potentially fraudulent phrases. The input is the content of the call, and the output is the fraud risk assessment result. Specifically, the voice data is converted into text and analyzed.

[0884] Step 4: Emotional state analysis and alert generation

[0885] The emotion engine analyzes the user's emotional state during a call in real time and generates an alert based on the fraud risk assessment results. The input is the call content and emotional data, and the output is an alert message. Specifically, if the user feels anxious, a warning message will be displayed on the device.

[0886] Step 5: Notify users

[0887] The device displays the message "You have received an anonymous call. Please contact your family or the police to confirm the details." The input is the fraud risk assessment result, and the output is the display of a warning message.

[0888] Fraudulent wording learning function

[0889] Processing Steps:

[0890] Step 1: Collect data on past fraud cases

[0891] The server collects past fraudulent email and message cases from a database. The input is past fraud case data, and the output is a training dataset.

[0892] Step 2: Train using machine learning algorithms

[0893] The server uses a machine learning algorithm to learn the fraudulent claims: the input is a training dataset and the output is a trained model.

[0894] Step 3: Detect new fraud methods

[0895] The server analyzes new fraudulent messages in real time and determines the likelihood of fraud based on a learning model. The input is new message data, and the output is a fraud risk assessment result and a warning message.

[0896] Step 4: Learning emotion data

[0897] The emotion engine learns from the user's emotion data, understands the user's unique emotional patterns, and more precisely assesses fraud risk. The input is the user's emotion data, and the output is an updated emotion model.

[0898] Example prompts to be input to the generative AI model

[0899] Detect fraudulent emails

[0900] "Assess the fraud risk of anonymous calls"

[0901] "Analyze the user's emotional state using an emotion engine."

[0902] (Application example 2)

[0903] 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."

[0904] In modern society, frauds that exploit email and social networking services are on the rise. Furthermore, there are a wide variety of frauds that take place over the phone, resulting in an increasing number of victims. These frauds pose a significant risk, especially for those with low information literacy, such as the elderly. Furthermore, systems that issue uniform warnings without considering the user's emotional state make it difficult to effectively avoid fraud. Furthermore, as fraud methods evolve daily, conventional systems have the problem of being unable to respond quickly to new fraud methods.

[0905] The identification processing 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 receiving email messages and messages from social networking services; means for analyzing the received messages to detect potentially fraudulent wording or unreliable URLs; means for displaying an alert when there is a possibility of fraud; means for responding to calls from anonymous or unknown phone numbers with an automated voice and analyzing the content of the call in real time; means for learning from past fraud cases and updating a model to adapt to new fraud methods; means for analyzing the user's emotional state and adjusting the strength of the warning based on the emotional score; and means for providing a natural language processing algorithm for managing fraud risks in real time. This enables early detection of fraud risks and effective warnings based on the user's emotional state.

[0906] An "email message" is a communication containing information such as text, images, and links that is sent and received electronically over the Internet.

[0907] A "social networking service" is an online platform that enables people to share information, communicate, and interact over the Internet.

[0908] "Potential fraud" refers to situations in which a message or call received may have been sent with the intent to deceive the user.

[0909] An "untrusted URL" is an internet address that is likely to link to a fraudulent or malicious site.

[0910] The "means for displaying an alert" refers to a device or software that has the function of displaying a warning message or notification on a terminal to alert the user.

[0911] "Private or unknown phone number" refers to a phone number that does not display caller number information or is not registered in the user's contact list.

[0912] "Automatic voice response means" refers to the ability of the system to automatically respond to incoming calls with a pre-recorded voice message.

[0913] "Means for analyzing call content in real time" refers to a device or software that has the function of converting voice during a call into text and instantly analyzing the content.

[0914] "Means for learning from past fraud cases" refers to the ability to collect and analyze data on previous fraud cases and use that information to generate models that can respond to new fraud methods.

[0915] "Means for adjusting the strength of the warning based on the emotional score" refers to a function that quantifies the user's emotional state and adjusts the strength and display method of the warning message according to that score.

[0916] A "natural language processing algorithm" is a computational method for computers to understand, interpret, and generate human language.

[0917] An embodiment of a system that realizes this application example will be described below. This system has the function of receiving email messages and messages from social networking services (SNS) and analyzing their contents to detect fraud risks. It also has the function of responding to calls from blocked or unknown phone numbers with an automated voice and analyzing the contents of the call in real time. It also has the function of analyzing the user's emotional state and adjusting the strength of the warning based on the emotional score. The specific configuration and operation of this system will be described below.

[0918] The server has a means for receiving email messages and SNS messages. It retrieves messages using an email client or SNS API. These messages are then analyzed using a natural language processing algorithm (e.g., BERT) to detect potentially fraudulent content and unreliable URLs. If there is a possibility of fraud based on the detection results, a warning message is displayed on the device. At this time, the user's emotional state is analyzed and the strength of the warning is adjusted according to the emotional score. A sentiment analysis library such as VADER is used to analyze the emotional state.

[0919] When the device receives a call from a blocked or unknown number, it responds with an automated voice and analyzes the call in real time. It converts the call into text using technologies like Google Speech-to-Text and then analyzes it with natural language processing algorithms. If there is a possibility of fraud, the device notifies the user of the call and issues appropriate warnings if necessary.

[0920] The server collects and learns from past fraud cases and updates the machine learning model to adapt to new fraud techniques. This makes it possible to adapt to new fraud techniques. For example, the server can learn phrases such as "Please transfer the money quickly" and detect similar phrases in the future.

[0921] For example, if a user receives an email that asks them to enter their bank account information, the server analyzes the message to detect possible fraud. If emotion analysis reveals that the user is under stress, a stronger warning message will be displayed on the device: "This email may be fraudulent. Please be careful."

[0922] Furthermore, if an anonymous call is received and the content of the call is detected as "Please transfer 1 million yen immediately," the user will be shown a message saying, "You have received an anonymous call. Please contact your family or the police to confirm the content."

[0923] An example of a prompt sentence might be given to the generative AI model: "Analyze new emails and display a warning if there is any potentially fraudulent content. Vary the strength of the warning depending on the user's emotional state."

[0924] In this way, the system enables early detection of fraud risks and effective warnings based on the user's emotional state.

[0925] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[0926] Step 1: Receiving email and SNS messages

[0927] The server receives user email and SNS messages through email clients and SNS APIs. The input is new emails and SNS messages, and the output is the body of the received messages. Specifically, the server periodically accesses the email server or SNS platform to retrieve unread messages.

[0928] Step 2: Parsing the message content

[0929] The server analyzes the received message content using a natural language processing (NLP) algorithm (e.g., BERT). The input is the message body, and the output is a judgment result (e.g., a fraud score) on whether the message is likely to be fraudulent. Specifically, the message body is tokenized and input into an NLP model to evaluate the likelihood of fraud.

[0930] Step 3: Sentiment Analysis

[0931] The server analyzes the user's emotional state using an emotion analysis library such as VADER. The input is the user's message content and voice data, and the output is an emotion score. Specifically, it extracts emotional features from the message and voice data, inputs them into an emotion analysis model, and obtains an emotion score.

[0932] Step 4: Viewing warnings

[0933] If a potential fraud is detected, the device displays a warning message to the user. The strength of this warning is adjusted based on the emotion score. The input is the fraud score and the emotion score, and the output is the warning message. Specifically, the device compares the fraud score with the emotion score, generates an appropriate warning message, and displays it on the device's display.

[0934] Step 5: Auto-answering calls

[0935] When the device receives a call from a blocked or unknown phone number, it responds with an automated voice message. The input is the incoming call data, and the output is an automated voice response. Specifically, it plays a pre-recorded voice message and records the call in real time.

[0936] Step 6: Analyzing the call

[0937] The server converts the recorded conversations into text in real time and analyzes them using an NLP algorithm. The input is the call audio data, and the output is a judgment result on whether there is a possibility of fraud. Specifically, the audio data is converted into text using Google Speech-to-Text, and the text is then input into the NLP model for analysis.

[0938] Step 7: View call alerts

[0939] If a possible fraud is detected, the terminal displays a warning message about the call content to the user. The input is the analysis result, and the output is the call warning message. Specifically, the warning message is generated based on the analysis result and displayed on the terminal.

[0940] Step 8: Learn and update your scam language

[0941] The server periodically updates the machine learning model using past fraud cases. The input is past fraud data, and the output is an updated learning model. Specifically, the server uses collected fraud data to retrain the model so that it can adapt to new fraud patterns.

[0942] The above are the specific processing steps of the system that realizes the application example. At each step, appropriate data processing and calculations are performed based on the input data, and warnings and notifications are generated for the user.

[0943] 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.

[0944] 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.

[0945] 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.

[0946] [Third embodiment]

[0947] FIG. 5 shows an example of the configuration of a data processing system 310 according to the third embodiment.

[0948] 5, the data processing system 310 includes the data processing device 12 and a headset type terminal 314. An example of the data processing device 12 is a server.

[0949] 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).

[0950] 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.

[0951] 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.

[0952] 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).

[0953] 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.

[0954] 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.

[0955] 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.

[0956] 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.

[0957] 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.

[0958] 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."

[0959] The present invention relates to a system for reducing the risk of special fraud occurring through email, social networking services, and telephone. This system is realized by a program having the following functions.

[0960] Email and SNS fraud detection function

[0961] When a device receives a new email or message from a social networking service, the server analyzes the message's content. Specifically, the server uses natural language processing algorithms to analyze the message text and detect potentially fraudulent phrases and unreliable URLs. If the server determines that a message is potentially fraudulent, it displays an alert message on the device to warn the user.

[0962] Examples:

[0963] 1. Your device will receive an email asking you to enter your bank account information.

[0964] 2. The server analyzes the text and detects possible fraud.

[0965] 3. The device notifies the user, "This email may be fraudulent. Please be careful."

[0966] Telephone fraud response features

[0967] When the device receives a call from a blocked or unknown number, the device will initiate an automated voice response. AI will analyze the call content in real time, and if there is a possibility of fraud, the server will notify the user of the call content and request confirmation of the call.

[0968] Examples:

[0969] 1. The device receives an anonymous call.

[0970] 2. The device will respond with an automated voice saying, "This call is being recorded for security reasons. Please wait as this may be a scam."

[0971] 3. The AI ​​analyzes the content of the call and detects the phrase "I urgently need money."

[0972] 4. The server detects possible fraud and

[0973] 5. The device notifies the user, "You have received an anonymous call. Please contact your family or the police to confirm the details."

[0974] Fraudulent wording learning function

[0975] The server collects data on past fraud cases and uses machine learning algorithms to learn fraudulent phrases. This allows the model to be updated to handle new fraud methods and enable early detection of unknown fraud methods. If a new fraud method is detected, the device will display a warning to the user based on that information.

[0976] Examples:

[0977] 1. The server analyzes past fraudulent emails and learns the common phrase "Please transfer the money quickly."

[0978] 2. The server adds the new fraud technique to the learning model.

[0979] 3. The server detects a new fraudulent email that says, "Click this link to receive a special discount."

[0980] 4. The device notifies the user, "This email may be a new scam. Do not click on the link."

[0981] As described above, the system according to the present invention can detect fraudulent acts via email, social media, and telephone, and can prevent fraud damage by issuing a warning to the user in advance, thereby effectively reducing fraud damage.

[0982] The processing flow will be explained below.

[0983] Email and SNS fraud detection function

[0984] Step 1:

[0985] Your device receives a new email or SNS message.

[0986] The device will notify you when a new message arrives in your inbox.

[0987] Step 2:

[0988] Gets the contents of the message received by the server.

[0989] The server stores the message data sent from the terminal in an analysis buffer.

[0990] Step 3:

[0991] The server begins parsing the message.

[0992] The server uses natural language processing (NLP) algorithms to tokenize and pattern match the message text.

[0993] Step 4:

[0994] The server assesses the likelihood of fraud.

[0995] The server determines the likelihood that the message content is fraudulent based on a registered list of fraudulent phrases and a reliability score.

[0996] Step 5:

[0997] Determines whether the server generates an alert.

[0998] The server decides to generate an alert message if the likelihood of fraud exceeds a threshold.

[0999] Step 6:

[1000] The device displays an alert to the user.

[1001] The device will display a pop-up warning message saying, "This email may be fraudulent. Please be careful."

[1002] Telephone fraud response features

[1003] Step 1:

[1004] Your device receives a call from a blocked or unknown number.

[1005] The device retrieves the caller information and matches it with an existing contact database.

[1006] Step 2:

[1007] The device will start an automatic voice response.

[1008] The terminal will begin the process of playing a pre-configured auto-answer message.

[1009] Step 3:

[1010] AI analyzes the content of calls in real time.

[1011] AI uses voice recognition technology to convert calls into text data, which is then analyzed using NLP algorithms.

[1012] Step 4:

[1013] The server assesses the likelihood of fraud.

[1014] Based on the analysis results, the server scores the likelihood that the call content is fraudulent.

[1015] Step 5:

[1016] The server asks the user to confirm the call.

[1017] If the server determines that there is a high possibility of fraud, it sends a warning message to the user's terminal.

[1018] Step 6:

[1019] The terminal displays a call confirmation to the user.

[1020] The device will display a message saying, "You have received an anonymous call. Please contact your family or the police to confirm the details."

[1021] Fraudulent wording learning function

[1022] Step 1:

[1023] The server collects historical fraud data.

[1024] The server stores past fraudulent emails and phone call details in a database.

[1025] Step 2:

[1026] The server learns the fraudulent language.

[1027] The server uses machine learning algorithms to analyze and extract common fraud patterns from the collected data.

[1028] Step 3:

[1029] The server updates the model.

[1030] The server adds and updates the learning model with newly discovered fraud techniques and patterns.

[1031] Step 4:

[1032] The server detects new fraud methods in real time.

[1033] The server applies the updated model to detect new fraud techniques from the latest data.

[1034] Step 5:

[1035] The terminal alerts the user based on new fraud patterns.

[1036] The device will display a notification saying, "This email may be a new scam. Do not click on any links."

[1037] The above are the processing steps in a specific embodiment of the present invention.

[1038] Example 1

[1039] 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."

[1040] Conventional systems have had difficulty effectively reducing the risk of fraud via email, social media, and telephone. In particular, they were unable to adapt to new fraud methods, increasing the likelihood of fraud victims occurring. Furthermore, users often felt uneasy because they were unable to analyze call content in real time to detect potential fraud.

[1041] 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.

[1042] In this invention, the server includes: means for receiving email messages and messages from social networking services; means for analyzing the received messages to detect potentially fraudulent wording and unreliable URLs; means for displaying an alert when there is a possibility of fraud; means for responding to calls from anonymous or unknown phone numbers with an automated voice and analyzing the content of the call in real time; means for learning from past fraud cases and updating the model to respond to new fraud techniques; and means for displaying a warning to the user when there is a possibility of fraud among the above means; means for analyzing the content of messages and calls in real time using a natural language processing algorithm to detect possible fraud; and means for detecting new fraud techniques using a generative AI model based on the analysis results. This makes it possible to detect fraudulent acts via email, social networking services, and telephone calls early and prevent fraud damage by issuing advance warnings to users.

[1043] "Email message" refers to the text information of an email sent or received over the Internet.

[1044] A "social networking service" is a platform that allows people to share information and communicate online.

[1045] "Means of receiving" refers to the technology and equipment used to receive email messages and messages from social networking services.

[1046] "Means for analysis" refers to the technology and equipment used to analyze the content of received messages and find specific patterns or characteristics.

[1047] "Potentially fraudulent language and unreliable URLs" refers to keywords associated with fraudulent activities and unreliable web links.

[1048] "Means for displaying an alert" refers to the technology and devices used to display a warning message to the user when a potential fraud is detected.

[1049] "Anonymous or unknown phone numbers" are phone numbers that cannot be identified by the caller and that have not been previously registered.

[1050] "Automatic voice response means" means technology and equipment that automatically plays a voice message in response to an incoming call without human intervention.

[1051] "Means for analyzing call content in real time" refers to technology and equipment that allows for immediate analysis of the content of a call while it is being made.

[1052] "Past fraud cases" refers to specific cases and data of fraudulent acts that have occurred in the past.

[1053] "Model updating methods" refers to the techniques and methods used to keep fraud detection algorithms up to date based on new information and data.

[1054] A "natural language processing algorithm" is an algorithm that allows a computer to analyze and understand human language.

[1055] A "generative AI model" is an artificial intelligence model that uses machine learning techniques to generate new patterns and knowledge from data.

[1056] "Means for displaying a warning to the user" refers to technologies and devices that provide visual or audio warnings to the user when possible fraud is detected.

[1057] The present invention relates to a system for reducing the risk of special frauds occurring through email, social networking services (SNS), and telephone. This system includes functions for receiving email and SNS messages, analyzing the received messages, determining the possibility of fraud, displaying alerts, responding to telephone fraud, and learning fraudulent phrases.

[1058] Email and SNS fraud detection function

[1059] The server receives emails and SNS messages using an internet-connected device (smartphone, PC, tablet, etc.). The received message is sent to the server via a secure communication protocol (e.g., HTTPS). The server analyzes the content of the message using a natural language processing algorithm (e.g., spaCy, NLTK) to detect specific keywords (e.g., 'bank account information', 'transfer') and unreliable URLs. If it determines that there is a possibility of fraud, an alert message is generated and sent to the device. The device notifies the user, "This email may be fraudulent. Please be careful."

[1060] Specific examples

[1061] When the terminal receives an email saying "Please enter your bank account information," it sends the received information to the server.

[1062] The server analyzes the email and detects possible fraud.

[1063] The device will notify the user, "This email may be fraudulent. Please be careful."

[1064] Prompt Sentence Examples

[1065] Here's a fraud warning message for when you receive an email that says "enter your bank account details."

[1066] Telephone fraud response features

[1067] When the device receives a call from an anonymous or unknown phone number, an automated voice response begins, saying, "This call is being recorded for safety reasons. Please wait as this may be a scam." The content of the call is analyzed in real time by AI, and the results are sent to the server. The server analyzes the call text, and if it detects phrases that are likely to be fraudulent (e.g., 'I urgently need money'), it sends a notification to the user requesting that they check the content of the call. The device notifies the user, "You have received an anonymous call. Please contact your family or the police to confirm the content."

[1068] Specific examples

[1069] When the device receives an anonymous call, an automated voice responds saying, "This call is being recorded for safety reasons. Please wait as this may be a scam."

[1070] AI analyzes the content of calls in real time and detects phrases such as "I urgently need money."

[1071] The server determines the possibility of fraud, and the device notifies the user, "You have received an anonymous call. Please contact your family or the police to confirm the details."

[1072] Prompt Sentence Examples

[1073] Display a warning message if an anonymous caller claims to be in urgent need of money.

[1074] Fraudulent wording learning function

[1075] The server collects data on past fraud cases and analyzes it using machine learning algorithms (e.g., TensorFlow, Scikit-learn). It learns characteristic patterns and keywords from the analyzed data and updates the model to respond to new fraud methods. The server detects new fraud methods and sends a notification to the device. The device displays a warning to the user saying, "This email may be a new scam. Do not click on the link."

[1076] Specific examples

[1077] The server analyzes past fraudulent emails and learns the phrase "Please transfer the money quickly."

[1078] The server adds new fraud techniques to the learning model.

[1079] The server now detects the phrase "Click this link to receive a special discount."

[1080] The device will notify the user, "This email may be a new scam. Do not click on the link."

[1081] Prompt Sentence Examples

[1082] A fraud warning message for those who receive emails that say, "Click this link to receive a special discount."

[1083] This system can detect fraudulent activities via email, social media, and telephone at an early stage and warn users in advance, preventing fraud damage before it occurs. This will significantly reduce the damage caused by fraud.

[1084] The flow of the identification process in the first embodiment will be described with reference to FIG.

[1085] Email and SNS fraud detection function

[1086] Processing Steps

[1087] Step 1:

[1088] The device receives new emails or SNS messages and saves their contents to local storage. The input is the received message, and the output is the saved message data.

[1089] Step 2:

[1090] The terminal sends the received message to the server using a secure communication protocol (e.g. HTTPS). The input is the stored message data, and the output is the message data sent to the server.

[1091] Step 3:

[1092] The server analyzes the message data it receives. Specifically, it uses a natural language processing library (e.g., spaCy, NLTK) to tokenize the message, tag parts of speech, and analyze context. The input is the message data sent to the server, and the output is the analysis results.

[1093] Step 4:

[1094] The server determines the likelihood of fraud based on the analysis results. Criteria for determining fraud include specific keywords (e.g., 'bank account information', 'transfer') and untrustworthy URLs. The input is the analysis results, and the output is an assessment of the likelihood of fraud.

[1095] Step 5:

[1096] If the server detects a possibility of fraud, it generates an alert message and sends it to the terminal. The input is the evaluation result of the possibility of fraud, and the output is the generated alert message.

[1097] Step 6:

[1098] The terminal notifies the user of the alert message. The input is the generated alert message, and the output is the warning message displayed to the user.

[1099] Telephone fraud response features

[1100] Processing Steps

[1101] Step 1:

[1102] The terminal receives a call from a blocked or unknown phone number. The input is the call from the blocked or unknown phone number, and the output is the acceptance of the call.

[1103] Step 2:

[1104] The terminal starts an automated voice response, saying, "This call is being recorded for security reasons. Please wait as this may be a scam." The input is the acceptance of the call, and the output is the automated voice response.

[1105] Step 3:

[1106] AI analyzes the contents of calls in real time. Specifically, it converts the contents of calls into text using the Google Speech-to-Text API and performs keyword analysis. The input is the contents of the calls, and the output is the analysis results.

[1107] Step 4:

[1108] The server determines the likelihood of fraud based on the analysis results. Fraud criteria include specific phrases (e.g., 'I urgently need money'). The input is the analysis results, and the output is an assessment of the likelihood of fraud.

[1109] Step 5:

[1110] If the server detects the possibility of fraud, it generates a notification requesting the user to confirm the contents of the call and sends it to the terminal. The input is the result of the evaluation of the possibility of fraud, and the output is the generated notification.

[1111] Step 6:

[1112] The device sends a notification to the user, warning them that "An anonymous call has been made. Please contact your family or the police to confirm the details." The input is the generated notification, and the output is the warning message that is displayed to the user.

[1113] Fraudulent wording learning function

[1114] Processing Steps

[1115] Step 1:

[1116] The server collects past fraudulent email and call data from the Internet and internal databases. The input is past fraud case data, and the output is the collected data.

[1117] Step 2:

[1118] The server uses machine learning algorithms (e.g., TensorFlow, Scikit-learn) to analyze the collected data and learn specific patterns or keywords. The input is the collected data, and the output is an updated learning model.

[1119] Step 3:

[1120] The server adds new fraud techniques to the learning model. The input is specific patterns or keywords, and the output is an updated learning model.

[1121] Step 4:

[1122] The server analyzes newly received emails and messages to detect new fraud methods. The input is the new email or message, and the output is the evaluation result of the likelihood of fraud.

[1123] Step 5:

[1124] If the device detects a possible scam, it notifies the user, "This email contains a new potential scam. Do not click on the link." The input is the result of the evaluation of the possible scam, and the output is a warning message that is displayed to the user.

[1125] (Application example 1)

[1126] 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."

[1127] In modern society, fraudulent activities via email, social networking services (SNS), and telephone are on the rise, and many people are falling victim to these scams. Elderly people and those unfamiliar with the Internet have particular difficulty distinguishing between fraudulent messages and phone calls, making it easier for the damage to spread. There is a need for technology that can effectively detect such fraudulent activities and issue early warnings to users.

[1128] 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.

[1129] In this invention, the server includes means for receiving email messages and messages from social networking services, means for analyzing the received messages to detect potentially fraudulent wording and unreliable URLs, means for displaying an alert when there is a possibility of fraud, means for responding to calls from withheld or unknown phone numbers with an automated voice and analyzing the content of the call in real time, means for learning from past fraud cases and updating the model to deal with new fraud methods, and means for visually warning users of the risk of fraud using a display device worn by the user. This makes it easier for users to recognize the risk of fraud in real time and prevents them from becoming victims of fraud.

[1130] "Mail message" refers to information, including text and attachments, sent and received via email.

[1131] A "social networking service" is an online platform that enables people to share information and deepen their interactions via the Internet.

[1132] A "server" is a computer system that processes and provides data over a network.

[1133] "Means for receiving messages" refers to technology that has the function of receiving emails and SNS messages on a user's device.

[1134] "Means for analyzing" refers to technology for analyzing the content of received messages to assess fraud risk.

[1135] "Wording" refers to a sentence or phrase, and refers to specific text that may be fraudulent.

[1136] "URL" stands for Uniform Resource Locator and is a description that indicates the address to a web page or online resource.

[1137] "Means for displaying an alert" refers to technology that displays a warning or alert message to the user.

[1138] "Calls from blocked or unknown numbers" refers to calls from phones that do not display or are not registered and cannot be identified.

[1139] "Means for responding with automated voice" refers to technology in which a system automatically generates and responds with a voice message.

[1140] "Means for analyzing call content in real time" refers to technology that instantly converts voice data during a call into text, analyzes the content, and assesses the risk of fraud.

[1141] "Past fraud cases" refers to data on fraud methods and cases that have been reported to date.

[1142] "Means of learning and updating models to adapt to new fraud techniques" refers to a technique that uses machine learning algorithms to learn new fraud patterns and update existing detection models.

[1143] A "display device" refers to hardware for visually presenting information, such as smart contact lenses or smart glasses.

[1144] "Means for visually warning of fraud risk" refers to technology that uses a visual display device worn by the user to display a warning about fraud risk.

[1145] The present invention relates to a system for reducing the risk of special frauds occurring through email, social networking services (SNS), and telephone. This system is realized by having the following main functions.

[1146] 1. Email and SNS fraud detection function

[1147] When a device receives a new email or message from a social networking service, the server analyzes the message's content. Specifically, the server uses a natural language processing library (e.g., Spacy) and a machine learning model (e.g., a model saved by scikit-learn) to analyze the message text and detect potentially fraudulent phrases or unreliable URLs. If any are detected, a warning message is displayed on the device in real time to alert the user.

[1148] 2. Telephone fraud response function

[1149] When the device receives a call from a blocked or unknown number, the device will initiate an automated voice response. AI will analyze the call content in real time, and if there is a possibility of fraud, the server will notify the user of the call content and request confirmation of the call.

[1150] 3. Fraudulent wording learning function

[1151] The server collects data on past fraud cases and uses machine learning algorithms to learn fraudulent phrases. This allows the model to be updated to handle new fraud methods and enable early detection of unknown fraud methods. If a new fraud method is detected, the device will display a warning to the user based on that information.

[1152] 4. Warning display function on smart contact lenses

[1153] The system also includes a function that visually warns users of fraud risks using display devices such as smart contact lenses or smart glasses. If the server detects a risk of fraud while the user is checking email or social media messages, a visual warning message saying "Possible fraud" will be displayed on the display device.

[1154] Hardware / software configuration used

[1155] Server: Analyzes messages, runs AI models, and updates learning models.

[1156] Device: Receives emails and SNS messages, records phone calls, and provides automated voice responses.

[1157] Natural language processing libraries such as Spacy can be used to analyze message content.

[1158] Machine learning models: Assess fraud risk using models saved in Scikit-learn.

[1159] Display devices: Use smart contact lenses or smart glasses to display warning messages.

[1160] Specific examples

[1161] For example, if a user receives an email requesting "Please enter your bank account information," the server analyzes the text and detects the possibility of fraud. The device (such as a smartphone) then displays a warning message saying, "This email may be fraudulent. Please be careful." At the same time, if the user is wearing smart contact lenses, the lenses will also display a visual warning saying, "This email may be fraudulent."

[1162] Prompt Sentence Examples

[1163] An example prompt for new messages would look something like this:

[1164] "A new message has arrived. Please analyze its contents."

[1165] As a result, this system can effectively detect fraudulent activities via email, social media, and telephone, and can prevent fraud damage by issuing advance warnings to users, thereby effectively reducing fraud damage.

[1166] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[1167] Step 1:

[1168] The device receives a new email or message from a social networking service. The received message data is provided as input. The device sends this message to the server.

[1169] Step 2:

[1170] The server analyzes the content of the received message. The input is the text data of the received message, and the output is the analysis result. The server uses a natural language processing library (e.g., Spacy) to tokenize the message text and detect potentially fraudulent phrases and unreliable URLs.

[1171] Step 3:

[1172] The server determines the likelihood of fraud based on the analysis results. The input is the message analysis results, and the output is the fraud risk assessment result. The fraud risk is assessed using a machine learning model trained on past fraud cases.

[1173] Step 4:

[1174] If there is a possibility of fraud, the server sends a warning to the terminal and display device. The input is the fraud risk assessment result (if the fraud risk is high), and the output is a warning message to be displayed to the user. The server generates a warning message such as "Possible fraud" and sends it to the terminal and display device such as a smart contact lens.

[1175] Step 5:

[1176] When a user receives a call from a blocked or unknown phone number, the device starts an automated voice response. The input is the incoming call information, and the output is an automated voice message. The device generates an automated voice message informing the user, "This call is being recorded for safety reasons. Please wait as this may be a scam."

[1177] Step 6:

[1178] The server analyzes the call content in real time. The input is the voice data of the call, and the output is the text conversion and analysis results of the call content. The server uses AI to convert the call content into text and analyze it.

[1179] Step 7:

[1180] The server evaluates the possibility of fraud based on the analysis results of the call content. The input is the text data of the call content and the analysis results, and the output is the fraud risk assessment result.

[1181] Step 8:

[1182] If there is a possibility of fraud, the server sends a warning to the terminal and display device and requests the user to confirm the call. The input is the fraud risk assessment result (if the fraud risk is high), and the output is a warning message to be displayed to the user and a request to confirm the content of the call. The server generates a warning message saying "There has been an anonymous call. Please consult with your family or the police to confirm the content," and sends it to the terminal and display device.

[1183] Step 9:

[1184] The server collects data on past fraud cases and uses a machine learning algorithm to learn fraudulent phrases. The input is data on past fraud cases, and the output is an updated fraud detection model. If a new fraud technique is detected, it is added to the learning model and reflected in the next analysis.

[1185] Through the above processing steps, this system can effectively detect fraudulent activities via email, social media, and telephone, and can prevent fraudulent acts by issuing advance warnings to users.

[1186] 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.

[1187] The present invention combines an emotion engine with a system for reducing the risk of special frauds occurring through email, social networking services, and telephone calls, enabling appropriate responses to be taken according to the user's emotional state. This system is realized by a program with the following functions:

[1188] Email and SNS fraud detection function

[1189] When a device receives a new email or message from a social networking service, the server retrieves and analyzes the message content. Specifically, the server uses a natural language processing (NLP) algorithm to analyze the message text and detect potentially fraudulent language or unreliable URLs. If the possibility of fraud is assessed, an alert message is displayed on the device to warn the user. At this time, the emotion engine recognizes the user's emotional state, and if the user is in a state of stress, a stronger alert is displayed.

[1190] Examples:

[1191] 1. Your device will receive an email asking you to enter your bank account information.

[1192] 2. The server analyzes the text and detects possible fraud.

[1193] 3. The emotion engine analyzes the user's emotional state and sets the appropriate alert intensity.

[1194] 4. The device will pop up a warning message saying, "This email may be fraudulent. Please be careful."

[1195] Telephone fraud response features

[1196] When a device receives a call from a blocked or unknown number, it will initiate an automated voice response. AI analyzes the call content in real time, and if there is a possibility of fraud, the server will notify the user of the call content and request confirmation. During this process, an emotion engine analyzes the user's emotional state in real time and adjusts the intensity and timing of the warning based on emotional changes.

[1197] Examples:

[1198] 1. The device receives an anonymous call.

[1199] 2. The device will respond with an automated voice saying, "This call is being recorded for security reasons. Please wait as this may be a scam."

[1200] 3. The AI ​​analyzes the content of the call and detects the phrase "I urgently need money."

[1201] 4. The emotion engine analyzes the user's emotional state during the call, and if the user feels anxious, the server generates an alert with appropriate intensity.

[1202] 5. The device will display the message, "You have received an anonymous call. Please contact your family or the police to confirm the details."

[1203] Fraudulent wording learning function

[1204] The server collects data on past fraud cases and uses machine learning algorithms to learn fraudulent phrases. This allows the model to be updated to handle new fraud methods, enabling early detection of unknown fraud methods. The emotion engine also learns from the user's emotional data, allowing it to understand the user's unique emotional patterns and more precisely determine the likelihood of fraud. If a new fraud method is detected, the device will display a warning to the user based on that information.

[1205] Examples:

[1206] 1. The server analyzes past fraudulent emails and learns the common phrase "Please transfer the money quickly."

[1207] 2. The server adds the new fraud technique to the learning model.

[1208] 3. The server detects a new fraudulent email that says, "Click this link to receive a special discount."

[1209] 4. The emotion engine uses user emotional data to more precisely assess the likelihood of fraud.

[1210] 5. Your device will display a notification saying, "This email may be a new scam. Do not click on any links."

[1211] As described above, the system according to the present invention can detect fraudulent acts via email, social media, and telephone, and take appropriate measures according to the user's emotional state to prevent fraud damage before it occurs, thereby effectively reducing fraud damage.

[1212] The processing flow will be explained below.

[1213] Email and SNS fraud detection function

[1214] Step 1:

[1215] Your device receives a new email or SNS message.

[1216] The device will notify you when a new message arrives in your inbox.

[1217] Step 2:

[1218] Gets the contents of the message received by the server.

[1219] The server stores the message data sent from the terminal in an analysis buffer.

[1220] Step 3:

[1221] The server begins parsing the message.

[1222] The server uses natural language processing (NLP) algorithms to tokenize and pattern match the message text.

[1223] Step 4:

[1224] An emotion engine analyzes the user's emotional state.

[1225] The emotion engine uses the user's facial expressions and behavioral data to assess their emotional state in real time.

[1226] Step 5:

[1227] The server assesses the likelihood of fraud.

[1228] The server determines the likelihood that the message content is fraudulent based on a registered list of fraudulent phrases and a reliability score.

[1229] Step 6:

[1230] Determines whether the server generates an alert.

[1231] If the possibility of fraud exceeds a threshold, the server generates an alert message based on the evaluation results of the emotion engine.

[1232] Step 7:

[1233] The device displays an alert to the user.

[1234] The device will display a pop-up warning message saying, "This email may be fraudulent. Please be careful."

[1235] Telephone fraud response features

[1236] Step 1:

[1237] Your device receives a call from a blocked or unknown number.

[1238] The device retrieves the caller information and matches it with an existing contact database.

[1239] Step 2:

[1240] The device will start an automatic voice response.

[1241] The terminal will begin the process of playing a pre-configured auto-answer message.

[1242] Step 3:

[1243] AI analyzes the content of calls in real time.

[1244] AI uses voice recognition technology to convert calls into text data, which is then analyzed using NLP algorithms.

[1245] Step 4:

[1246] The emotion engine analyzes the user's emotional state in real time.

[1247] The emotion engine analyzes the user's tone of voice and vocabulary to assess emotional fluctuations.

[1248] Step 5:

[1249] The server assesses the likelihood of fraud.

[1250] Based on the analysis results, the server scores the likelihood that the call content is fraudulent.

[1251] Step 6:

[1252] The server asks the user to confirm the call.

[1253] If the server determines that there is a high possibility of fraud, it generates an alert of appropriate strength based on the evaluation results of the emotion engine and sends it to the device.

[1254] Step 7:

[1255] The terminal displays a call confirmation to the user.

[1256] The device will display a message saying, "You have received an anonymous call. Please contact your family or the police to confirm the details."

[1257] Fraudulent wording learning function

[1258] Step 1:

[1259] The server collects historical fraud data.

[1260] The server stores past fraudulent emails and phone call details in a database.

[1261] Step 2:

[1262] The server learns the fraudulent language.

[1263] The server uses machine learning algorithms to analyze and extract common fraud patterns from the collected data.

[1264] Step 3:

[1265] The server updates the model.

[1266] The server adds and updates the learning model with newly discovered fraud techniques and patterns.

[1267] Step 4:

[1268] The emotion engine learns the user's emotion data.

[1269] The emotion engine learns the user's unique emotional patterns to more precisely assess the likelihood of fraud.

[1270] Step 5:

[1271] The server detects new fraud methods in real time.

[1272] The server applies the updated model to detect new fraud techniques from the latest data.

[1273] Step 6:

[1274] The terminal alerts the user based on new fraud patterns.

[1275] The device will display a notification saying, "This email may be a new scam. Do not click on any links."

[1276] The above are the processing steps in a specific embodiment of the present invention.

[1277] Example 2

[1278] 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."

[1279] In recent years, fraudulent activities via email and social networking services have been increasing, and many users have fallen victim to them. Fraudulent activities using anonymous and unknown phone numbers are also rampant. As fraudsters become more sophisticated, it is becoming increasingly difficult for users to distinguish fraudulent messages and calls, making it difficult to prevent damage before it occurs. In order to reduce the damage caused by fraud, a system that can detect possible fraud in real time and warn users appropriately is needed.

[1280] The specific processing by the specific processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means.

[1281] In this invention, the server includes means for receiving email messages and messages from social networking services, means for analyzing the received messages to detect potentially fraudulent wording and unreliable URLs, means for displaying an alert when there is a possibility of fraud, means for responding to calls from anonymous or unknown phone numbers with an automated voice and analyzing the content of the call in real time, means for learning from past fraud cases and updating the model to deal with new fraud methods, means for analyzing the user's emotional state and adjusting the strength and timing of the alert, and means for generating a warning message for the user based on the results of the analysis of the call content and messages. This makes it possible to detect fraudulent acts in real time and provide an optimal warning according to the user's emotional state.

[1282] An "email message" is an electronic document sent or received over the Internet that may contain text, images, links, etc.

[1283] "Social networking service" refers to a platform that enables users to interact and share information with other users on the Internet.

[1284] "Means of receiving" refers to the technical methods and equipment used to receive emails and messages into the user's terminal.

[1285] "Means for analyzing" refers to technical methods or devices that process the content of received messages or calls and analyze their content.

[1286] "Potentially fraudulent language" refers to words or phrases used with the intent to fraudulently obtain money or personal information.

[1287] "Untrusted URLs" refer to web links that are likely to be fraudulent or phishing.

[1288] "Means for displaying an alert" refers to a display device or method for issuing a warning or alert to the user.

[1289] "Hidden or unknown phone number" refers to a phone number where the caller's phone number is not displayed or is not registered by the user.

[1290] "Automated voice response means" means any technology or method that uses pre-recorded messages to answer telephone calls.

[1291] "Means of real-time analysis" refers to technologies and devices that process data the moment it is generated and output the results immediately.

[1292] "Past fraud cases" refers to data that records fraud cases and methods that have occurred in the past.

[1293] "Means for updating the model" refers to methods for updating machine learning algorithms and databases with the latest information.

[1294] "User's emotional state" refers to the psychological state or mood exhibited by a user at a particular moment.

[1295] "Means for adjusting the intensity and timing of alerts" refers to technologies and methods for appropriately changing the content and timing of warnings based on the user's emotional state.

[1296] "Means for generating a warning message" refers to a technique or method for creating and displaying a message to alert the user.

[1297] The present invention relates to a system for reducing the risk of fraudulent activity via email, social networking service (SNS) messages, and telephone calls. This system is implemented by a program with multiple functions. Each function is described in detail below, along with specific examples.

[1298] Email and SNS fraud detection function

[1299] When a device receives a new email or social media message, the server retrieves the message content and begins analyzing it. The server uses natural language processing (NLP) algorithms to detect potentially fraudulent content and unreliable URLs. Specific software that can be used is the popular NLP library "spaCy" or "NLTK." If a potential fraud is detected, an alert message is displayed on the device. An emotion engine is used to analyze the user's emotional state and adjust the intensity and content of the alert.

[1300] Examples:

[1301] 1. Your device will receive an email asking you to enter your bank account information.

[1302] 2. The server analyzes the text and detects possible fraud.

[1303] 3. The emotion engine analyzes the user's emotional state and sets the appropriate alert intensity.

[1304] 4. The device will pop up a warning message saying, "This email may be fraudulent. Please be careful."

[1305] Telephone fraud response features

[1306] When the device receives a call from a blocked or unknown phone number, it initiates an automated voice response. AI analyzes the call content in real time, and if there is a possibility of fraud, the server notifies the user of the call content and requests confirmation of the call. During this process, an emotion engine analyzes the user's emotional state in real time and adjusts the strength and timing of the warning based on emotional changes. The voice recognition software "Google Speech-to-Text API" and "Amazon Transcribe" can be used to analyze the call content.

[1307] Examples:

[1308] 1. The device receives an anonymous call.

[1309] 2. The device will respond with an automated voice saying, "This call is being recorded for security reasons. Please wait as this may be a scam."

[1310] 3. The AI ​​analyzes the content of the call and detects the phrase "I urgently need money."

[1311] 4. The emotion engine analyzes the user's emotional state during the call, and if the user feels anxious, the server generates an alert with appropriate intensity.

[1312] 5. The device will display the message, "You have received an anonymous call. Please contact your family or the police to confirm the details."

[1313] Fraudulent wording learning function

[1314] The server collects data on past fraud cases and uses machine learning algorithms to learn fraudulent phrases. This allows the model to be updated to deal with new fraud methods. Technologies used include machine learning libraries such as "scikit-learn" and "TensorFlow." This function makes it possible to detect unknown fraud methods early and take preventative measures. The emotion engine also learns from users' emotional data, understanding their unique emotional patterns to more precisely determine the likelihood of fraud.

[1315] Examples:

[1316] 1. The server analyzes past fraudulent emails and learns the common phrase "Please transfer the money quickly."

[1317] 2. The server adds the new fraud technique to the learning model.

[1318] 3. The server detects a new fraudulent email that says, "Click this link to receive a special discount."

[1319] 4. The emotion engine uses user emotional data to more precisely assess the likelihood of fraud.

[1320] 5. Your device will display a notification saying, "This email may be a new scam. Do not click on any links."

[1321] Example prompts to be input to the generative AI model

[1322] Detect fraudulent emails

[1323] "Assess the fraud risk of anonymous calls"

[1324] "Analyze the user's emotional state using an emotion engine."

[1325] The above is a specific embodiment of the system according to the present invention. The system of the present invention can detect fraudulent acts via email, social media, and telephone, and take appropriate action according to the user's emotional state, thereby preventing fraud damage before it occurs. This makes it possible to effectively reduce fraud damage.

[1326] The flow of the identification process in the second embodiment will be described with reference to FIG.

[1327] Email and SNS fraud detection function

[1328] Processing Steps:

[1329] Step 1: Receive an email or social media message

[1330] When a device receives a new email or SNS message, it temporarily stores the contents in memory. This process requires no user input; the new message is the input data. The output is the message content itself.

[1331] Step 2: Parsing the message content

[1332] The server retrieves the received message content and analyzes it using a natural language processing (NLP) algorithm. Specifically, the message text is the input data, and the NLP algorithm detects potentially fraudulent phrases and unreliable URLs and assigns a score. The output is a score indicating the likelihood of fraud and the analysis results.

[1333] Step 3: Assess the likelihood of fraud

[1334] The server evaluates the fraud probability score based on the analysis results. If the score exceeds a certain threshold, it is determined that there is a high probability of fraud. The input is the analysis result from the previous step, and the output is the fraud risk assessment result.

[1335] Step 4: Alert the user

[1336] The terminal receives the fraud risk assessment results, and the emotion engine analyzes the user's emotional state and adjusts the strength and content of the alert. The input is the fraud risk assessment results and the user's emotional data, and the output is a warning display as a pop-up message. Specifically, the warning message is displayed in a pop-up window.

[1337] Telephone fraud response features

[1338] Processing Steps:

[1339] Step 1: Call from a blocked or unknown number

[1340] The terminal receives a call from a blocked or unknown number. The input is the incoming call signal and the output is the initiation of the call connection.

[1341] Step 2: Automated voice response

[1342] The terminal plays an automated voice message saying, "This call is being recorded for security reasons. Please wait as this may be a scam." The input is the start signal for the call connection, and the output is the automated voice message.

[1343] Step 3: Real-time analysis of call content

[1344] The AI ​​on the server analyzes the content of the call in real time and detects potentially fraudulent phrases. The input is the content of the call, and the output is the fraud risk assessment result. Specifically, the voice data is converted into text and analyzed.

[1345] Step 4: Emotional state analysis and alert generation

[1346] The emotion engine analyzes the user's emotional state during a call in real time and generates an alert based on the fraud risk assessment results. The input is the call content and emotional data, and the output is an alert message. Specifically, if the user feels anxious, a warning message will be displayed on the device.

[1347] Step 5: Notify users

[1348] The device displays the message "You have received an anonymous call. Please contact your family or the police to confirm the details." The input is the fraud risk assessment result, and the output is the display of a warning message.

[1349] Fraudulent wording learning function

[1350] Processing Steps:

[1351] Step 1: Collect data on past fraud cases

[1352] The server collects past fraudulent email and message cases from a database. The input is past fraud case data, and the output is a training dataset.

[1353] Step 2: Train using machine learning algorithms

[1354] The server uses a machine learning algorithm to learn the fraudulent claims: the input is a training dataset and the output is a trained model.

[1355] Step 3: Detect new fraud methods

[1356] The server analyzes new fraudulent messages in real time and determines the likelihood of fraud based on a learning model. The input is new message data, and the output is a fraud risk assessment result and a warning message.

[1357] Step 4: Learning emotion data

[1358] The emotion engine learns from the user's emotion data, understands the user's unique emotional patterns, and more precisely assesses fraud risk. The input is the user's emotion data, and the output is an updated emotion model.

[1359] Example prompts to be input to the generative AI model

[1360] Detect fraudulent emails

[1361] "Assess the fraud risk of anonymous calls"

[1362] "Analyze the user's emotional state using an emotion engine."

[1363] (Application example 2)

[1364] 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."

[1365] In modern society, frauds that exploit email and social networking services are on the rise. Furthermore, there are a wide variety of frauds that take place over the phone, resulting in an increasing number of victims. These frauds pose a significant risk, especially for those with low information literacy, such as the elderly. Furthermore, systems that issue uniform warnings without considering the user's emotional state make it difficult to effectively avoid fraud. Furthermore, as fraud methods evolve daily, conventional systems have the problem of being unable to respond quickly to new fraud methods.

[1366] The identification processing 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 receiving email messages and messages from social networking services; means for analyzing the received messages to detect potentially fraudulent wording or unreliable URLs; means for displaying an alert when there is a possibility of fraud; means for responding to calls from anonymous or unknown phone numbers with an automated voice and analyzing the content of the call in real time; means for learning from past fraud cases and updating a model to adapt to new fraud methods; means for analyzing the user's emotional state and adjusting the strength of the warning based on the emotional score; and means for providing a natural language processing algorithm for managing fraud risks in real time. This enables early detection of fraud risks and effective warnings based on the user's emotional state.

[1367] An "email message" is a communication containing information such as text, images, and links that is sent and received electronically over the Internet.

[1368] A "social networking service" is an online platform that enables people to share information, communicate, and interact over the Internet.

[1369] "Potential fraud" refers to situations in which a message or call received may have been sent with the intent to deceive the user.

[1370] An "untrusted URL" is an internet address that is likely to link to a fraudulent or malicious site.

[1371] The "means for displaying an alert" refers to a device or software that has the function of displaying a warning message or notification on a terminal to alert the user.

[1372] "Private or unknown phone number" refers to a phone number that does not display caller number information or is not registered in the user's contact list.

[1373] "Automatic voice response means" refers to the ability of the system to automatically respond to incoming calls with a pre-recorded voice message.

[1374] "Means for analyzing call content in real time" refers to a device or software that has the function of converting voice during a call into text and instantly analyzing the content.

[1375] "Means for learning from past fraud cases" refers to the ability to collect and analyze data on previous fraud cases and use that information to generate models that can respond to new fraud methods.

[1376] "Means for adjusting the strength of the warning based on the emotional score" refers to a function that quantifies the user's emotional state and adjusts the strength and display method of the warning message according to that score.

[1377] A "natural language processing algorithm" is a computational method for computers to understand, interpret, and generate human language.

[1378] An embodiment of a system that realizes this application example will be described below. This system has the function of receiving email messages and messages from social networking services (SNS) and analyzing their contents to detect fraud risks. It also has the function of responding to calls from blocked or unknown phone numbers with an automated voice and analyzing the contents of the call in real time. It also has the function of analyzing the user's emotional state and adjusting the strength of the warning based on the emotional score. The specific configuration and operation of this system will be described below.

[1379] The server has a means for receiving email messages and SNS messages. It retrieves messages using an email client or SNS API. These messages are then analyzed using a natural language processing algorithm (e.g., BERT) to detect potentially fraudulent content and unreliable URLs. If there is a possibility of fraud based on the detection results, a warning message is displayed on the device. At this time, the user's emotional state is analyzed and the strength of the warning is adjusted according to the emotional score. A sentiment analysis library such as VADER is used to analyze the emotional state.

[1380] When the device receives a call from a blocked or unknown number, it responds with an automated voice and analyzes the call in real time. It converts the call into text using technologies like Google Speech-to-Text and then analyzes it with natural language processing algorithms. If there is a possibility of fraud, the device notifies the user of the call and issues appropriate warnings if necessary.

[1381] The server collects and learns from past fraud cases and updates the machine learning model to adapt to new fraud techniques. This makes it possible to adapt to new fraud techniques. For example, the server can learn phrases such as "Please transfer the money quickly" and detect similar phrases in the future.

[1382] For example, if a user receives an email that asks them to enter their bank account information, the server analyzes the message to detect possible fraud. If emotion analysis reveals that the user is under stress, a stronger warning message will be displayed on the device: "This email may be fraudulent. Please be careful."

[1383] Furthermore, if an anonymous call is received and the content of the call is detected as "Please transfer 1 million yen immediately," the user will be shown a message saying, "You have received an anonymous call. Please contact your family or the police to confirm the content."

[1384] An example of a prompt sentence might be given to the generative AI model: "Analyze new emails and display a warning if there is any potentially fraudulent content. Vary the strength of the warning depending on the user's emotional state."

[1385] In this way, the system enables early detection of fraud risks and effective warnings based on the user's emotional state.

[1386] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[1387] Step 1: Receiving email and SNS messages

[1388] The server receives user email and SNS messages through email clients and SNS APIs. The input is new emails and SNS messages, and the output is the body of the received messages. Specifically, the server periodically accesses the email server or SNS platform to retrieve unread messages.

[1389] Step 2: Parsing the message content

[1390] The server analyzes the received message content using a natural language processing (NLP) algorithm (e.g., BERT). The input is the message body, and the output is a judgment result (e.g., a fraud score) on whether the message is likely to be fraudulent. Specifically, the message body is tokenized and input into an NLP model to evaluate the likelihood of fraud.

[1391] Step 3: Sentiment Analysis

[1392] The server analyzes the user's emotional state using an emotion analysis library such as VADER. The input is the user's message content and voice data, and the output is an emotion score. Specifically, it extracts emotional features from the message and voice data, inputs them into an emotion analysis model, and obtains an emotion score.

[1393] Step 4: Viewing warnings

[1394] If a potential fraud is detected, the device displays a warning message to the user. The strength of this warning is adjusted based on the emotion score. The input is the fraud score and the emotion score, and the output is the warning message. Specifically, the device compares the fraud score with the emotion score, generates an appropriate warning message, and displays it on the device's display.

[1395] Step 5: Auto-answering calls

[1396] When the device receives a call from a blocked or unknown phone number, it responds with an automated voice message. The input is the incoming call data, and the output is an automated voice response. Specifically, it plays a pre-recorded voice message and records the call in real time.

[1397] Step 6: Analyzing the call

[1398] The server converts the recorded conversations into text in real time and analyzes them using an NLP algorithm. The input is the call audio data, and the output is a judgment result on whether there is a possibility of fraud. Specifically, the audio data is converted into text using Google Speech-to-Text, and the text is then input into the NLP model for analysis.

[1399] Step 7: View call alerts

[1400] If a possible fraud is detected, the terminal displays a warning message about the call content to the user. The input is the analysis result, and the output is the call warning message. Specifically, the warning message is generated based on the analysis result and displayed on the terminal.

[1401] Step 8: Learn and update your scam language

[1402] The server periodically updates the machine learning model using past fraud cases. The input is past fraud data, and the output is an updated learning model. Specifically, the server uses collected fraud data to retrain the model so that it can adapt to new fraud patterns.

[1403] The above are the specific processing steps of the system that realizes the application example. At each step, appropriate data processing and calculations are performed based on the input data, and warnings and notifications are generated for the user.

[1404] 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.

[1405] 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.

[1406] 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.

[1407] [Fourth embodiment]

[1408] FIG. 7 shows an example of the configuration of a data processing system 410 according to the fourth embodiment.

[1409] 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.

[1410] 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).

[1411] 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.

[1412] 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.

[1413] 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).

[1414] 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.

[1415] 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.

[1416] 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.

[1417] 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.

[1418] 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.

[1419] 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.

[1420] 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."

[1421] The present invention relates to a system for reducing the risk of special fraud occurring through email, social networking services, and telephone. This system is realized by a program having the following functions.

[1422] Email and SNS fraud detection function

[1423] When a device receives a new email or message from a social networking service, the server analyzes the message's content. Specifically, the server uses natural language processing algorithms to analyze the message text and detect potentially fraudulent phrases and unreliable URLs. If the server determines that a message is potentially fraudulent, it displays an alert message on the device to warn the user.

[1424] Examples:

[1425] 1. Your device will receive an email asking you to enter your bank account information.

[1426] 2. The server analyzes the text and detects possible fraud.

[1427] 3. The device notifies the user, "This email may be fraudulent. Please be careful."

[1428] Telephone fraud response features

[1429] When the device receives a call from a blocked or unknown number, the device will initiate an automated voice response. AI will analyze the call content in real time, and if there is a possibility of fraud, the server will notify the user of the call content and request confirmation of the call.

[1430] Examples:

[1431] 1. The device receives an anonymous call.

[1432] 2. The device will respond with an automated voice saying, "This call is being recorded for security reasons. Please wait as this may be a scam."

[1433] 3. The AI ​​analyzes the content of the call and detects the phrase "I urgently need money."

[1434] 4. The server detects possible fraud and

[1435] 5. The device notifies the user, "You have received an anonymous call. Please contact your family or the police to confirm the details."

[1436] Fraudulent wording learning function

[1437] The server collects data on past fraud cases and uses machine learning algorithms to learn fraudulent phrases. This allows the model to be updated to handle new fraud methods and enable early detection of unknown fraud methods. If a new fraud method is detected, the device will display a warning to the user based on that information.

[1438] Examples:

[1439] 1. The server analyzes past fraudulent emails and learns the common phrase "Please transfer the money quickly."

[1440] 2. The server adds the new fraud technique to the learning model.

[1441] 3. The server detects a new fraudulent email that says, "Click this link to receive a special discount."

[1442] 4. The device notifies the user, "This email may be a new scam. Do not click on the link."

[1443] As described above, the system according to the present invention can detect fraudulent acts via email, social media, and telephone, and can prevent fraud damage by issuing a warning to the user in advance, thereby effectively reducing fraud damage.

[1444] The processing flow will be explained below.

[1445] Email and SNS fraud detection function

[1446] Step 1:

[1447] Your device receives a new email or SNS message.

[1448] The device will notify you when a new message arrives in your inbox.

[1449] Step 2:

[1450] Gets the contents of the message received by the server.

[1451] The server stores the message data sent from the terminal in an analysis buffer.

[1452] Step 3:

[1453] The server begins parsing the message.

[1454] The server uses natural language processing (NLP) algorithms to tokenize and pattern match the message text.

[1455] Step 4:

[1456] The server assesses the likelihood of fraud.

[1457] The server determines the likelihood that the message content is fraudulent based on a registered list of fraudulent phrases and a reliability score.

[1458] Step 5:

[1459] Determines whether the server generates an alert.

[1460] The server decides to generate an alert message if the likelihood of fraud exceeds a threshold.

[1461] Step 6:

[1462] The device displays an alert to the user.

[1463] The device will display a pop-up warning message saying, "This email may be fraudulent. Please be careful."

[1464] Telephone fraud response features

[1465] Step 1:

[1466] Your device receives a call from a blocked or unknown number.

[1467] The device retrieves the caller information and matches it with an existing contact database.

[1468] Step 2:

[1469] The device will start an automatic voice response.

[1470] The terminal will begin the process of playing a pre-configured auto-answer message.

[1471] Step 3:

[1472] AI analyzes the content of calls in real time.

[1473] AI uses voice recognition technology to convert calls into text data, which is then analyzed using NLP algorithms.

[1474] Step 4:

[1475] The server assesses the likelihood of fraud.

[1476] Based on the analysis results, the server scores the likelihood that the call content is fraudulent.

[1477] Step 5:

[1478] The server asks the user to confirm the call.

[1479] If the server determines that there is a high possibility of fraud, it sends a warning message to the user's terminal.

[1480] Step 6:

[1481] The terminal displays a call confirmation to the user.

[1482] The device will display a message saying, "You have received an anonymous call. Please contact your family or the police to confirm the details."

[1483] Fraudulent wording learning function

[1484] Step 1:

[1485] The server collects historical fraud data.

[1486] The server stores past fraudulent emails and phone call details in a database.

[1487] Step 2:

[1488] The server learns the fraudulent language.

[1489] The server uses machine learning algorithms to analyze and extract common fraud patterns from the collected data.

[1490] Step 3:

[1491] The server updates the model.

[1492] The server adds and updates the learning model with newly discovered fraud techniques and patterns.

[1493] Step 4:

[1494] The server detects new fraud methods in real time.

[1495] The server applies the updated model to detect new fraud techniques from the latest data.

[1496] Step 5:

[1497] The terminal alerts the user based on new fraud patterns.

[1498] The device will display a notification saying, "This email may be a new scam. Do not click on any links."

[1499] The above are the processing steps in a specific embodiment of the present invention.

[1500] Example 1

[1501] 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."

[1502] Conventional systems have had difficulty effectively reducing the risk of fraud via email, social media, and telephone. In particular, they were unable to adapt to new fraud methods, increasing the likelihood of fraud victims occurring. Furthermore, users often felt uneasy because they were unable to analyze call content in real time to detect potential fraud.

[1503] 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.

[1504] In this invention, the server includes: means for receiving email messages and messages from social networking services; means for analyzing the received messages to detect potentially fraudulent wording and unreliable URLs; means for displaying an alert when there is a possibility of fraud; means for responding to calls from anonymous or unknown phone numbers with an automated voice and analyzing the content of the call in real time; means for learning from past fraud cases and updating the model to respond to new fraud techniques; and means for displaying a warning to the user when there is a possibility of fraud among the above means; means for analyzing the content of messages and calls in real time using a natural language processing algorithm to detect possible fraud; and means for detecting new fraud techniques using a generative AI model based on the analysis results. This makes it possible to detect fraudulent acts via email, social networking services, and telephone calls early and prevent fraud damage by issuing advance warnings to users.

[1505] "Email message" refers to the text information of an email sent or received over the Internet.

[1506] A "social networking service" is a platform that allows people to share information and communicate online.

[1507] "Means of receiving" refers to the technology and equipment used to receive email messages and messages from social networking services.

[1508] "Means for analysis" refers to the technology and equipment used to analyze the content of received messages and find specific patterns or characteristics.

[1509] "Potentially fraudulent language and unreliable URLs" refers to keywords associated with fraudulent activities and unreliable web links.

[1510] "Means for displaying an alert" refers to the technology and devices used to display a warning message to the user when a potential fraud is detected.

[1511] "Anonymous or unknown phone numbers" are phone numbers that cannot be identified by the caller and that have not been previously registered.

[1512] "Automatic voice response means" means technology and equipment that automatically plays a voice message in response to an incoming call without human intervention.

[1513] "Means for analyzing call content in real time" refers to technology and equipment that allows for immediate analysis of the content of a call while it is being made.

[1514] "Past fraud cases" refers to specific cases and data of fraudulent acts that have occurred in the past.

[1515] "Model updating methods" refers to the techniques and methods used to keep fraud detection algorithms up to date based on new information and data.

[1516] A "natural language processing algorithm" is an algorithm that allows a computer to analyze and understand human language.

[1517] A "generative AI model" is an artificial intelligence model that uses machine learning techniques to generate new patterns and knowledge from data.

[1518] "Means for displaying a warning to the user" refers to technologies and devices that provide visual or audio warnings to the user when possible fraud is detected.

[1519] The present invention relates to a system for reducing the risk of special frauds occurring through email, social networking services (SNS), and telephone. This system includes functions for receiving email and SNS messages, analyzing the received messages, determining the possibility of fraud, displaying alerts, responding to telephone fraud, and learning fraudulent phrases.

[1520] Email and SNS fraud detection function

[1521] The server receives emails and SNS messages using an internet-connected device (smartphone, PC, tablet, etc.). The received message is sent to the server via a secure communication protocol (e.g., HTTPS). The server analyzes the content of the message using a natural language processing algorithm (e.g., spaCy, NLTK) to detect specific keywords (e.g., 'bank account information', 'transfer') and unreliable URLs. If it determines that there is a possibility of fraud, an alert message is generated and sent to the device. The device notifies the user, "This email may be fraudulent. Please be careful."

[1522] Specific examples

[1523] When the terminal receives an email saying "Please enter your bank account information," it sends the received information to the server.

[1524] The server analyzes the email and detects possible fraud.

[1525] The device will notify the user, "This email may be fraudulent. Please be careful."

[1526] Prompt Sentence Examples

[1527] Here's a fraud warning message for when you receive an email that says "enter your bank account details."

[1528] Telephone fraud response features

[1529] When the device receives a call from an anonymous or unknown phone number, an automated voice response begins, saying, "This call is being recorded for safety reasons. Please wait as this may be a scam." The content of the call is analyzed in real time by AI, and the results are sent to the server. The server analyzes the call text, and if it detects phrases that are likely to be fraudulent (e.g., 'I urgently need money'), it sends a notification to the user requesting that they check the content of the call. The device notifies the user, "You have received an anonymous call. Please contact your family or the police to confirm the content."

[1530] Specific examples

[1531] When the device receives an anonymous call, an automated voice responds saying, "This call is being recorded for safety reasons. Please wait as this may be a scam."

[1532] AI analyzes the content of calls in real time and detects phrases such as "I urgently need money."

[1533] The server determines the possibility of fraud, and the device notifies the user, "You have received an anonymous call. Please contact your family or the police to confirm the details."

[1534] Prompt Sentence Examples

[1535] Display a warning message if an anonymous caller claims to be in urgent need of money.

[1536] Fraudulent wording learning function

[1537] The server collects data on past fraud cases and analyzes it using machine learning algorithms (e.g., TensorFlow, Scikit-learn). It learns characteristic patterns and keywords from the analyzed data and updates the model to respond to new fraud methods. The server detects new fraud methods and sends a notification to the device. The device displays a warning to the user saying, "This email may be a new scam. Do not click on the link."

[1538] Specific examples

[1539] The server analyzes past fraudulent emails and learns the phrase "Please transfer the money quickly."

[1540] The server adds new fraud techniques to the learning model.

[1541] The server now detects the phrase "Click this link to receive a special discount."

[1542] The device will notify the user, "This email may be a new scam. Do not click on the link."

[1543] Prompt Sentence Examples

[1544] A fraud warning message for those who receive emails that say, "Click this link to receive a special discount."

[1545] This system can detect fraudulent activities via email, social media, and telephone at an early stage and warn users in advance, preventing fraud damage before it occurs. This will significantly reduce the damage caused by fraud.

[1546] The flow of the identification process in the first embodiment will be described with reference to FIG.

[1547] Email and SNS fraud detection function

[1548] Processing Steps

[1549] Step 1:

[1550] The device receives new emails or SNS messages and saves their contents to local storage. The input is the received message, and the output is the saved message data.

[1551] Step 2:

[1552] The terminal sends the received message to the server using a secure communication protocol (e.g. HTTPS). The input is the stored message data, and the output is the message data sent to the server.

[1553] Step 3:

[1554] The server analyzes the message data it receives. Specifically, it uses a natural language processing library (e.g., spaCy, NLTK) to tokenize the message, tag parts of speech, and analyze context. The input is the message data sent to the server, and the output is the analysis results.

[1555] Step 4:

[1556] The server determines the likelihood of fraud based on the analysis results. Criteria for determining fraud include specific keywords (e.g., 'bank account information', 'transfer') and untrustworthy URLs. The input is the analysis results, and the output is an assessment of the likelihood of fraud.

[1557] Step 5:

[1558] If the server detects a possibility of fraud, it generates an alert message and sends it to the terminal. The input is the evaluation result of the possibility of fraud, and the output is the generated alert message.

[1559] Step 6:

[1560] The terminal notifies the user of the alert message. The input is the generated alert message, and the output is the warning message displayed to the user.

[1561] Telephone fraud response features

[1562] Processing Steps

[1563] Step 1:

[1564] The terminal receives a call from a blocked or unknown phone number. The input is the call from the blocked or unknown phone number, and the output is the acceptance of the call.

[1565] Step 2:

[1566] The terminal starts an automated voice response, saying, "This call is being recorded for security reasons. Please wait as this may be a scam." The input is the acceptance of the call, and the output is the automated voice response.

[1567] Step 3:

[1568] AI analyzes the contents of calls in real time. Specifically, it converts the contents of calls into text using the Google Speech-to-Text API and performs keyword analysis. The input is the contents of the calls, and the output is the analysis results.

[1569] Step 4:

[1570] The server determines the likelihood of fraud based on the analysis results. Fraud criteria include specific phrases (e.g., 'I urgently need money'). The input is the analysis results, and the output is an assessment of the likelihood of fraud.

[1571] Step 5:

[1572] If the server detects the possibility of fraud, it generates a notification requesting the user to confirm the contents of the call and sends it to the terminal. The input is the result of the evaluation of the possibility of fraud, and the output is the generated notification.

[1573] Step 6:

[1574] The device sends a notification to the user, warning them that "An anonymous call has been made. Please contact your family or the police to confirm the details." The input is the generated notification, and the output is the warning message that is displayed to the user.

[1575] Fraudulent wording learning function

[1576] Processing Steps

[1577] Step 1:

[1578] The server collects past fraudulent email and call data from the Internet and internal databases. The input is past fraud case data, and the output is the collected data.

[1579] Step 2:

[1580] The server uses machine learning algorithms (e.g., TensorFlow, Scikit-learn) to analyze the collected data and learn specific patterns or keywords. The input is the collected data, and the output is an updated learning model.

[1581] Step 3:

[1582] The server adds new fraud techniques to the learning model. The input is specific patterns or keywords, and the output is an updated learning model.

[1583] Step 4:

[1584] The server analyzes newly received emails and messages to detect new fraud methods. The input is the new email or message, and the output is the evaluation result of the likelihood of fraud.

[1585] Step 5:

[1586] If the device detects a possible scam, it notifies the user, "This email contains a new potential scam. Do not click on the link." The input is the result of the evaluation of the possible scam, and the output is a warning message that is displayed to the user.

[1587] (Application example 1)

[1588] 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."

[1589] In modern society, fraudulent activities via email, social networking services (SNS), and telephone are on the rise, and many people are falling victim to these scams. Elderly people and those unfamiliar with the Internet have particular difficulty distinguishing between fraudulent messages and phone calls, making it easier for the damage to spread. There is a need for technology that can effectively detect such fraudulent activities and issue early warnings to users.

[1590] 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.

[1591] In this invention, the server includes means for receiving email messages and messages from social networking services, means for analyzing the received messages to detect potentially fraudulent wording and unreliable URLs, means for displaying an alert when there is a possibility of fraud, means for responding to calls from withheld or unknown phone numbers with an automated voice and analyzing the content of the call in real time, means for learning from past fraud cases and updating the model to deal with new fraud methods, and means for visually warning users of the risk of fraud using a display device worn by the user. This makes it easier for users to recognize the risk of fraud in real time and prevents them from becoming victims of fraud.

[1592] "Mail message" refers to information, including text and attachments, sent and received via email.

[1593] A "social networking service" is an online platform that enables people to share information and deepen their interactions via the Internet.

[1594] A "server" is a computer system that processes and provides data over a network.

[1595] "Means for receiving messages" refers to technology that has the function of receiving emails and SNS messages on a user's device.

[1596] "Means for analyzing" refers to technology for analyzing the content of received messages to assess fraud risk.

[1597] "Wording" refers to a sentence or phrase, and refers to specific text that may be fraudulent.

[1598] "URL" stands for Uniform Resource Locator and is a description that indicates the address to a web page or online resource.

[1599] "Means for displaying an alert" refers to technology that displays a warning or alert message to the user.

[1600] "Calls from blocked or unknown numbers" refers to calls from phones that do not display or are not registered and cannot be identified.

[1601] "Means for responding with automated voice" refers to technology in which a system automatically generates and responds with a voice message.

[1602] "Means for analyzing call content in real time" refers to technology that instantly converts voice data during a call into text, analyzes the content, and assesses the risk of fraud.

[1603] "Past fraud cases" refers to data on fraud methods and cases that have been reported to date.

[1604] "Means of learning and updating models to adapt to new fraud techniques" refers to a technique that uses machine learning algorithms to learn new fraud patterns and update existing detection models.

[1605] A "display device" refers to hardware for visually presenting information, such as smart contact lenses or smart glasses.

[1606] "Means for visually warning of fraud risk" refers to technology that uses a visual display device worn by the user to display a warning about fraud risk.

[1607] The present invention relates to a system for reducing the risk of special frauds occurring through email, social networking services (SNS), and telephone. This system is realized by having the following main functions.

[1608] 1. Email and SNS fraud detection function

[1609] When a device receives a new email or message from a social networking service, the server analyzes the message's content. Specifically, the server uses a natural language processing library (e.g., Spacy) and a machine learning model (e.g., a model saved by scikit-learn) to analyze the message text and detect potentially fraudulent phrases or unreliable URLs. If any are detected, a warning message is displayed on the device in real time to alert the user.

[1610] 2. Telephone fraud response function

[1611] When the device receives a call from a blocked or unknown number, the device will initiate an automated voice response. AI will analyze the call content in real time, and if there is a possibility of fraud, the server will notify the user of the call content and request confirmation of the call.

[1612] 3. Fraudulent wording learning function

[1613] The server collects data on past fraud cases and uses machine learning algorithms to learn fraudulent phrases. This allows the model to be updated to handle new fraud methods and enable early detection of unknown fraud methods. If a new fraud method is detected, the device will display a warning to the user based on that information.

[1614] 4. Warning display function on smart contact lenses

[1615] The system also includes a function that visually warns users of fraud risks using display devices such as smart contact lenses or smart glasses. If the server detects a risk of fraud while the user is checking email or social media messages, a visual warning message saying "Possible fraud" will be displayed on the display device.

[1616] Hardware / software configuration used

[1617] Server: Analyzes messages, runs AI models, and updates learning models.

[1618] Device: Receives emails and SNS messages, records phone calls, and provides automated voice responses.

[1619] Natural language processing libraries such as Spacy can be used to analyze message content.

[1620] Machine learning models: Assess fraud risk using models saved in Scikit-learn.

[1621] Display devices: Use smart contact lenses or smart glasses to display warning messages.

[1622] Specific examples

[1623] For example, if a user receives an email requesting "Please enter your bank account information," the server analyzes the text and detects the possibility of fraud. The device (such as a smartphone) then displays a warning message saying, "This email may be fraudulent. Please be careful." At the same time, if the user is wearing smart contact lenses, the lenses will also display a visual warning saying, "This email may be fraudulent."

[1624] Prompt Sentence Examples

[1625] An example prompt for new messages would look something like this:

[1626] "A new message has arrived. Please analyze its contents."

[1627] As a result, this system can effectively detect fraudulent activities via email, social media, and telephone, and can prevent fraud damage by issuing advance warnings to users, thereby effectively reducing fraud damage.

[1628] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[1629] Step 1:

[1630] The device receives a new email or message from a social networking service. The received message data is provided as input. The device sends this message to the server.

[1631] Step 2:

[1632] The server analyzes the content of the received message. The input is the text data of the received message, and the output is the analysis result. The server uses a natural language processing library (e.g., Spacy) to tokenize the message text and detect potentially fraudulent phrases and unreliable URLs.

[1633] Step 3:

[1634] The server determines the likelihood of fraud based on the analysis results. The input is the message analysis results, and the output is the fraud risk assessment result. The fraud risk is assessed using a machine learning model trained on past fraud cases.

[1635] Step 4:

[1636] If there is a possibility of fraud, the server sends a warning to the terminal and display device. The input is the fraud risk assessment result (if the fraud risk is high), and the output is a warning message to be displayed to the user. The server generates a warning message such as "Possible fraud" and sends it to the terminal and display device such as a smart contact lens.

[1637] Step 5:

[1638] When a user receives a call from a blocked or unknown phone number, the device starts an automated voice response. The input is the incoming call information, and the output is an automated voice message. The device generates an automated voice message informing the user, "This call is being recorded for safety reasons. Please wait as this may be a scam."

[1639] Step 6:

[1640] The server analyzes the call content in real time. The input is the voice data of the call, and the output is the text conversion and analysis results of the call content. The server uses AI to convert the call content into text and analyze it.

[1641] Step 7:

[1642] The server evaluates the possibility of fraud based on the analysis results of the call content. The input is the text data of the call content and the analysis results, and the output is the fraud risk assessment result.

[1643] Step 8:

[1644] If there is a possibility of fraud, the server sends a warning to the terminal and display device and requests the user to confirm the call. The input is the fraud risk assessment result (if the fraud risk is high), and the output is a warning message to be displayed to the user and a request to confirm the content of the call. The server generates a warning message saying "There has been an anonymous call. Please consult with your family or the police to confirm the content," and sends it to the terminal and display device.

[1645] Step 9:

[1646] The server collects data on past fraud cases and uses a machine learning algorithm to learn fraudulent phrases. The input is data on past fraud cases, and the output is an updated fraud detection model. If a new fraud technique is detected, it is added to the learning model and reflected in the next analysis.

[1647] Through the above processing steps, this system can effectively detect fraudulent activities via email, social media, and telephone, and can prevent fraudulent acts by issuing advance warnings to users.

[1648] 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.

[1649] The present invention combines an emotion engine with a system for reducing the risk of special frauds occurring through email, social networking services, and telephone calls, enabling appropriate responses to be taken according to the user's emotional state. This system is realized by a program with the following functions:

[1650] Email and SNS fraud detection function

[1651] When a device receives a new email or message from a social networking service, the server retrieves and analyzes the message content. Specifically, the server uses a natural language processing (NLP) algorithm to analyze the message text and detect potentially fraudulent language or unreliable URLs. If the possibility of fraud is assessed, an alert message is displayed on the device to warn the user. At this time, the emotion engine recognizes the user's emotional state, and if the user is in a state of stress, a stronger alert is displayed.

[1652] Examples:

[1653] 1. Your device will receive an email asking you to enter your bank account information.

[1654] 2. The server analyzes the text and detects possible fraud.

[1655] 3. The emotion engine analyzes the user's emotional state and sets the appropriate alert intensity.

[1656] 4. The device will pop up a warning message saying, "This email may be fraudulent. Please be careful."

[1657] Telephone fraud response features

[1658] When a device receives a call from a blocked or unknown number, it will initiate an automated voice response. AI analyzes the call content in real time, and if there is a possibility of fraud, the server will notify the user of the call content and request confirmation. During this process, an emotion engine analyzes the user's emotional state in real time and adjusts the intensity and timing of the warning based on emotional changes.

[1659] Examples:

[1660] 1. The device receives an anonymous call.

[1661] 2. The device will respond with an automated voice saying, "This call is being recorded for security reasons. Please wait as this may be a scam."

[1662] 3. The AI ​​analyzes the content of the call and detects the phrase "I urgently need money."

[1663] 4. The emotion engine analyzes the user's emotional state during the call, and if the user feels anxious, the server generates an alert with appropriate intensity.

[1664] 5. The device will display the message, "You have received an anonymous call. Please contact your family or the police to confirm the details."

[1665] Fraudulent wording learning function

[1666] The server collects data on past fraud cases and uses machine learning algorithms to learn fraudulent phrases. This allows the model to be updated to handle new fraud methods, enabling early detection of unknown fraud methods. The emotion engine also learns from the user's emotional data, allowing it to understand the user's unique emotional patterns and more precisely determine the likelihood of fraud. If a new fraud method is detected, the device will display a warning to the user based on that information.

[1667] Examples:

[1668] 1. The server analyzes past fraudulent emails and learns the common phrase "Please transfer the money quickly."

[1669] 2. The server adds the new fraud technique to the learning model.

[1670] 3. The server detects a new fraudulent email that says, "Click this link to receive a special discount."

[1671] 4. The emotion engine uses user emotional data to more precisely assess the likelihood of fraud.

[1672] 5. Your device will display a notification saying, "This email may be a new scam. Do not click on any links."

[1673] As described above, the system according to the present invention can detect fraudulent acts via email, social media, and telephone, and take appropriate measures according to the user's emotional state to prevent fraud damage before it occurs, thereby effectively reducing fraud damage.

[1674] The processing flow will be explained below.

[1675] Email and SNS fraud detection function

[1676] Step 1:

[1677] Your device receives a new email or SNS message.

[1678] The device will notify you when a new message arrives in your inbox.

[1679] Step 2:

[1680] Gets the contents of the message received by the server.

[1681] The server stores the message data sent from the terminal in an analysis buffer.

[1682] Step 3:

[1683] The server begins parsing the message.

[1684] The server uses natural language processing (NLP) algorithms to tokenize and pattern match the message text.

[1685] Step 4:

[1686] An emotion engine analyzes the user's emotional state.

[1687] The emotion engine uses the user's facial expressions and behavioral data to assess their emotional state in real time.

[1688] Step 5:

[1689] The server assesses the likelihood of fraud.

[1690] The server determines the likelihood that the message content is fraudulent based on a registered list of fraudulent phrases and a reliability score.

[1691] Step 6:

[1692] Determines whether the server generates an alert.

[1693] If the possibility of fraud exceeds a threshold, the server generates an alert message based on the evaluation results of the emotion engine.

[1694] Step 7:

[1695] The device displays an alert to the user.

[1696] The device will display a pop-up warning message saying, "This email may be fraudulent. Please be careful."

[1697] Telephone fraud response features

[1698] Step 1:

[1699] Your device receives a call from a blocked or unknown number.

[1700] The device retrieves the caller information and matches it with an existing contact database.

[1701] Step 2:

[1702] The device will start an automatic voice response.

[1703] The terminal will begin the process of playing a pre-configured auto-answer message.

[1704] Step 3:

[1705] AI analyzes the content of calls in real time.

[1706] AI uses voice recognition technology to convert calls into text data, which is then analyzed using NLP algorithms.

[1707] Step 4:

[1708] The emotion engine analyzes the user's emotional state in real time.

[1709] The emotion engine analyzes the user's tone of voice and vocabulary to assess emotional fluctuations.

[1710] Step 5:

[1711] The server assesses the likelihood of fraud.

[1712] Based on the analysis results, the server scores the likelihood that the call content is fraudulent.

[1713] Step 6:

[1714] The server asks the user to confirm the call.

[1715] If the server determines that there is a high possibility of fraud, it generates an alert of appropriate strength based on the evaluation results of the emotion engine and sends it to the device.

[1716] Step 7:

[1717] The terminal displays a call confirmation to the user.

[1718] The device will display a message saying, "You have received an anonymous call. Please contact your family or the police to confirm the details."

[1719] Fraudulent wording learning function

[1720] Step 1:

[1721] The server collects historical fraud data.

[1722] The server stores past fraudulent emails and phone call details in a database.

[1723] Step 2:

[1724] The server learns the fraudulent language.

[1725] The server uses machine learning algorithms to analyze and extract common fraud patterns from the collected data.

[1726] Step 3:

[1727] The server updates the model.

[1728] The server adds and updates the learning model with newly discovered fraud techniques and patterns.

[1729] Step 4:

[1730] The emotion engine learns the user's emotion data.

[1731] The emotion engine learns the user's unique emotional patterns to more precisely assess the likelihood of fraud.

[1732] Step 5:

[1733] The server detects new fraud methods in real time.

[1734] The server applies the updated model to detect new fraud techniques from the latest data.

[1735] Step 6:

[1736] The terminal alerts the user based on new fraud patterns.

[1737] The device will display a notification saying, "This email may be a new scam. Do not click on any links."

[1738] The above are the processing steps in a specific embodiment of the present invention.

[1739] Example 2

[1740] 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."

[1741] In recent years, fraudulent activities via email and social networking services have been increasing, and many users have fallen victim to them. Fraudulent activities using anonymous and unknown phone numbers are also rampant. As fraudsters become more sophisticated, it is becoming increasingly difficult for users to distinguish fraudulent messages and calls, making it difficult to prevent damage before it occurs. In order to reduce the damage caused by fraud, a system that can detect possible fraud in real time and warn users appropriately is needed.

[1742] The specific processing by the specific processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means.

[1743] In this invention, the server includes means for receiving email messages and messages from social networking services, means for analyzing the received messages to detect potentially fraudulent wording and unreliable URLs, means for displaying an alert when there is a possibility of fraud, means for responding to calls from anonymous or unknown phone numbers with an automated voice and analyzing the content of the call in real time, means for learning from past fraud cases and updating the model to deal with new fraud methods, means for analyzing the user's emotional state and adjusting the strength and timing of the alert, and means for generating a warning message for the user based on the results of the analysis of the call content and messages. This makes it possible to detect fraudulent acts in real time and provide an optimal warning according to the user's emotional state.

[1744] An "email message" is an electronic document sent or received over the Internet that may contain text, images, links, etc.

[1745] "Social networking service" refers to a platform that enables users to interact and share information with other users on the Internet.

[1746] "Means of receiving" refers to the technical methods and equipment used to receive emails and messages into the user's terminal.

[1747] "Means for analyzing" refers to technical methods or devices that process the content of received messages or calls and analyze their content.

[1748] "Potentially fraudulent language" refers to words or phrases used with the intent to fraudulently obtain money or personal information.

[1749] "Untrusted URLs" refer to web links that are likely to be fraudulent or phishing.

[1750] "Means for displaying an alert" refers to a display device or method for issuing a warning or alert to the user.

[1751] "Hidden or unknown phone number" refers to a phone number where the caller's phone number is not displayed or is not registered by the user.

[1752] "Automated voice response means" means any technology or method that uses pre-recorded messages to answer telephone calls.

[1753] "Means of real-time analysis" refers to technologies and devices that process data the moment it is generated and output the results immediately.

[1754] "Past fraud cases" refers to data that records fraud cases and methods that have occurred in the past.

[1755] "Means for updating the model" refers to methods for updating machine learning algorithms and databases with the latest information.

[1756] "User's emotional state" refers to the psychological state or mood exhibited by a user at a particular moment.

[1757] "Means for adjusting the intensity and timing of alerts" refers to technologies and methods for appropriately changing the content and timing of warnings based on the user's emotional state.

[1758] "Means for generating a warning message" refers to a technique or method for creating and displaying a message to alert the user.

[1759] The present invention relates to a system for reducing the risk of fraudulent activity via email, social networking service (SNS) messages, and telephone calls. This system is implemented by a program with multiple functions. Each function is described in detail below, along with specific examples.

[1760] Email and SNS fraud detection function

[1761] When a device receives a new email or social media message, the server retrieves the message content and begins analyzing it. The server uses natural language processing (NLP) algorithms to detect potentially fraudulent content and unreliable URLs. Specific software that can be used is the popular NLP library "spaCy" or "NLTK." If a potential fraud is detected, an alert message is displayed on the device. An emotion engine is used to analyze the user's emotional state and adjust the intensity and content of the alert.

[1762] Examples:

[1763] 1. Your device will receive an email asking you to enter your bank account information.

[1764] 2. The server analyzes the text and detects possible fraud.

[1765] 3. The emotion engine analyzes the user's emotional state and sets the appropriate alert intensity.

[1766] 4. The device will pop up a warning message saying, "This email may be fraudulent. Please be careful."

[1767] Telephone fraud response features

[1768] When the device receives a call from a blocked or unknown phone number, it initiates an automated voice response. AI analyzes the call content in real time, and if there is a possibility of fraud, the server notifies the user of the call content and requests confirmation of the call. During this process, an emotion engine analyzes the user's emotional state in real time and adjusts the strength and timing of the warning based on emotional changes. The voice recognition software "Google Speech-to-Text API" and "Amazon Transcribe" can be used to analyze the call content.

[1769] Examples:

[1770] 1. The device receives an anonymous call.

[1771] 2. The device will respond with an automated voice saying, "This call is being recorded for security reasons. Please wait as this may be a scam."

[1772] 3. The AI ​​analyzes the content of the call and detects the phrase "I urgently need money."

[1773] 4. The emotion engine analyzes the user's emotional state during the call, and if the user feels anxious, the server generates an alert with appropriate intensity.

[1774] 5. The device will display the message, "You have received an anonymous call. Please contact your family or the police to confirm the details."

[1775] Fraudulent wording learning function

[1776] The server collects data on past fraud cases and uses machine learning algorithms to learn fraudulent phrases. This allows the model to be updated to deal with new fraud methods. Technologies used include machine learning libraries such as "scikit-learn" and "TensorFlow." This function makes it possible to detect unknown fraud methods early and take preventative measures. The emotion engine also learns from users' emotional data, understanding their unique emotional patterns to more precisely determine the likelihood of fraud.

[1777] Examples:

[1778] 1. The server analyzes past fraudulent emails and learns the common phrase "Please transfer the money quickly."

[1779] 2. The server adds the new fraud technique to the learning model.

[1780] 3. The server detects a new fraudulent email that says, "Click this link to receive a special discount."

[1781] 4. The emotion engine uses user emotional data to more precisely assess the likelihood of fraud.

[1782] 5. Your device will display a notification saying, "This email may be a new scam. Do not click on any links."

[1783] Example prompts to be input to the generative AI model

[1784] Detect fraudulent emails

[1785] "Assess the fraud risk of anonymous calls"

[1786] "Analyze the user's emotional state using an emotion engine."

[1787] The above is a specific embodiment of the system according to the present invention. The system of the present invention can detect fraudulent acts via email, social media, and telephone, and take appropriate action according to the user's emotional state, thereby preventing fraud damage before it occurs. This makes it possible to effectively reduce fraud damage.

[1788] The flow of the identification process in the second embodiment will be described with reference to FIG.

[1789] Email and SNS fraud detection function

[1790] Processing Steps:

[1791] Step 1: Receive an email or social media message

[1792] When a device receives a new email or SNS message, it temporarily stores the contents in memory. This process requires no user input; the new message is the input data. The output is the message content itself.

[1793] Step 2: Parsing the message content

[1794] The server retrieves the received message content and analyzes it using a natural language processing (NLP) algorithm. Specifically, the message text is the input data, and the NLP algorithm detects potentially fraudulent phrases and unreliable URLs and assigns a score. The output is a score indicating the likelihood of fraud and the analysis results.

[1795] Step 3: Assess the likelihood of fraud

[1796] The server evaluates the fraud probability score based on the analysis results. If the score exceeds a certain threshold, it is determined that there is a high probability of fraud. The input is the analysis result from the previous step, and the output is the fraud risk assessment result.

[1797] Step 4: Alert the user

[1798] The terminal receives the fraud risk assessment results, and the emotion engine analyzes the user's emotional state and adjusts the strength and content of the alert. The input is the fraud risk assessment results and the user's emotional data, and the output is a warning display as a pop-up message. Specifically, the warning message is displayed in a pop-up window.

[1799] Telephone fraud response features

[1800] Processing Steps:

[1801] Step 1: Call from a blocked or unknown number

[1802] The terminal receives a call from a blocked or unknown number. The input is the incoming call signal and the output is the initiation of the call connection.

[1803] Step 2: Automated voice response

[1804] The terminal plays an automated voice message saying, "This call is being recorded for security reasons. Please wait as this may be a scam." The input is the start signal for the call connection, and the output is the automated voice message.

[1805] Step 3: Real-time analysis of call content

[1806] The AI ​​on the server analyzes the content of the call in real time and detects potentially fraudulent phrases. The input is the content of the call, and the output is the fraud risk assessment result. Specifically, the voice data is converted into text and analyzed.

[1807] Step 4: Emotional state analysis and alert generation

[1808] The emotion engine analyzes the user's emotional state during a call in real time and generates an alert based on the fraud risk assessment results. The input is the call content and emotional data, and the output is an alert message. Specifically, if the user feels anxious, a warning message will be displayed on the device.

[1809] Step 5: Notify users

[1810] The device displays the message "You have received an anonymous call. Please contact your family or the police to confirm the details." The input is the fraud risk assessment result, and the output is the display of a warning message.

[1811] Fraudulent wording learning function

[1812] Processing Steps:

[1813] Step 1: Collect data on past fraud cases

[1814] The server collects past fraudulent email and message cases from a database. The input is past fraud case data, and the output is a training dataset.

[1815] Step 2: Train using machine learning algorithms

[1816] The server uses a machine learning algorithm to learn the fraudulent claims: the input is a training dataset and the output is a trained model.

[1817] Step 3: Detect new fraud methods

[1818] The server analyzes new fraudulent messages in real time and determines the likelihood of fraud based on a learning model. The input is new message data, and the output is a fraud risk assessment result and a warning message.

[1819] Step 4: Learning emotion data

[1820] The emotion engine learns from the user's emotion data, understands the user's unique emotional patterns, and more precisely assesses fraud risk. The input is the user's emotion data, and the output is an updated emotion model.

[1821] Example prompts to be input to the generative AI model

[1822] Detect fraudulent emails

[1823] "Assess the fraud risk of anonymous calls"

[1824] "Analyze the user's emotional state using an emotion engine."

[1825] (Application example 2)

[1826] 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."

[1827] In modern society, frauds that exploit email and social networking services are on the rise. Furthermore, there are a wide variety of frauds that take place over the phone, resulting in an increasing number of victims. These frauds pose a significant risk, especially for those with low information literacy, such as the elderly. Furthermore, systems that issue uniform warnings without considering the user's emotional state make it difficult to effectively avoid fraud. Furthermore, as fraud methods evolve daily, conventional systems have the problem of being unable to respond quickly to new fraud methods.

[1828] The identification processing 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 receiving email messages and messages from social networking services; means for analyzing the received messages to detect potentially fraudulent wording or unreliable URLs; means for displaying an alert when there is a possibility of fraud; means for responding to calls from anonymous or unknown phone numbers with an automated voice and analyzing the content of the call in real time; means for learning from past fraud cases and updating a model to adapt to new fraud methods; means for analyzing the user's emotional state and adjusting the strength of the warning based on the emotional score; and means for providing a natural language processing algorithm for managing fraud risks in real time. This enables early detection of fraud risks and effective warnings based on the user's emotional state.

[1829] An "email message" is a communication containing information such as text, images, and links that is sent and received electronically over the Internet.

[1830] A "social networking service" is an online platform that enables people to share information, communicate, and interact over the Internet.

[1831] "Potential fraud" refers to situations in which a message or call received may have been sent with the intent to deceive the user.

[1832] An "untrusted URL" is an internet address that is likely to link to a fraudulent or malicious site.

[1833] The "means for displaying an alert" refers to a device or software that has the function of displaying a warning message or notification on a terminal to alert the user.

[1834] "Private or unknown phone number" refers to a phone number that does not display caller number information or is not registered in the user's contact list.

[1835] "Automatic voice response means" refers to the ability of the system to automatically respond to incoming calls with a pre-recorded voice message.

[1836] "Means for analyzing call content in real time" refers to a device or software that has the function of converting voice during a call into text and instantly analyzing the content.

[1837] "Means for learning from past fraud cases" refers to the ability to collect and analyze data on previous fraud cases and use that information to generate models that can respond to new fraud methods.

[1838] "Means for adjusting the strength of the warning based on the emotional score" refers to a function that quantifies the user's emotional state and adjusts the strength and display method of the warning message according to that score.

[1839] A "natural language processing algorithm" is a computational method for computers to understand, interpret, and generate human language.

[1840] An embodiment of a system that realizes this application example will be described below. This system has the function of receiving email messages and messages from social networking services (SNS) and analyzing their contents to detect fraud risks. It also has the function of responding to calls from blocked or unknown phone numbers with an automated voice and analyzing the contents of the call in real time. It also has the function of analyzing the user's emotional state and adjusting the strength of the warning based on the emotional score. The specific configuration and operation of this system will be described below.

[1841] The server has a means for receiving email messages and SNS messages. It retrieves messages using an email client or SNS API. These messages are then analyzed using a natural language processing algorithm (e.g., BERT) to detect potentially fraudulent content and unreliable URLs. If there is a possibility of fraud based on the detection results, a warning message is displayed on the device. At this time, the user's emotional state is analyzed and the strength of the warning is adjusted according to the emotional score. A sentiment analysis library such as VADER is used to analyze the emotional state.

[1842] When the device receives a call from a blocked or unknown number, it responds with an automated voice and analyzes the call in real time. It converts the call into text using technologies like Google Speech-to-Text and then analyzes it with natural language processing algorithms. If there is a possibility of fraud, the device notifies the user of the call and issues appropriate warnings if necessary.

[1843] The server collects and learns from past fraud cases and updates the machine learning model to adapt to new fraud techniques. This makes it possible to adapt to new fraud techniques. For example, the server can learn phrases such as "Please transfer the money quickly" and detect similar phrases in the future.

[1844] For example, if a user receives an email that asks them to enter their bank account information, the server analyzes the message to detect possible fraud. If emotion analysis reveals that the user is under stress, a stronger warning message will be displayed on the device: "This email may be fraudulent. Please be careful."

[1845] Furthermore, if an anonymous call is received and the content of the call is detected as "Please transfer 1 million yen immediately," the user will be shown a message saying, "You have received an anonymous call. Please contact your family or the police to confirm the content."

[1846] An example of a prompt sentence might be given to the generative AI model: "Analyze new emails and display a warning if there is any potentially fraudulent content. Vary the strength of the warning depending on the user's emotional state."

[1847] In this way, the system enables early detection of fraud risks and effective warnings based on the user's emotional state.

[1848] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[1849] Step 1: Receiving email and SNS messages

[1850] The server receives user email and SNS messages through email clients and SNS APIs. The input is new emails and SNS messages, and the output is the body of the received messages. Specifically, the server periodically accesses the email server or SNS platform to retrieve unread messages.

[1851] Step 2: Parsing the message content

[1852] The server analyzes the received message content using a natural language processing (NLP) algorithm (e.g., BERT). The input is the message body, and the output is a judgment result (e.g., a fraud score) on whether the message is likely to be fraudulent. Specifically, the message body is tokenized and input into an NLP model to evaluate the likelihood of fraud.

[1853] Step 3: Sentiment Analysis

[1854] The server analyzes the user's emotional state using an emotion analysis library such as VADER. The input is the user's message content and voice data, and the output is an emotion score. Specifically, it extracts emotional features from the message and voice data, inputs them into an emotion analysis model, and obtains an emotion score.

[1855] Step 4: Viewing warnings

[1856] If a potential fraud is detected, the device displays a warning message to the user. The strength of this warning is adjusted based on the emotion score. The input is the fraud score and the emotion score, and the output is the warning message. Specifically, the device compares the fraud score with the emotion score, generates an appropriate warning message, and displays it on the device's display.

[1857] Step 5: Auto-answering calls

[1858] When the device receives a call from a blocked or unknown phone number, it responds with an automated voice message. The input is the incoming call data, and the output is an automated voice response. Specifically, it plays a pre-recorded voice message and records the call in real time.

[1859] Step 6: Analyzing the call

[1860] The server converts the recorded conversations into text in real time and analyzes them using an NLP algorithm. The input is the call audio data, and the output is a judgment result on whether there is a possibility of fraud. Specifically, the audio data is converted into text using Google Speech-to-Text, and the text is then input into the NLP model for analysis.

[1861] Step 7: View call alerts

[1862] If a possible fraud is detected, the terminal displays a warning message about the call content to the user. The input is the analysis result, and the output is the call warning message. Specifically, the warning message is generated based on the analysis result and displayed on the terminal.

[1863] Step 8: Learn and update your scam language

[1864] The server periodically updates the machine learning model using past fraud cases. The input is past fraud data, and the output is an updated learning model. Specifically, the server uses collected fraud data to retrain the model so that it can adapt to new fraud patterns.

[1865] The above are the specific processing steps of the system that realizes the application example. At each step, appropriate data processing and calculations are performed based on the input data, and warnings and notifications are generated for the user.

[1866] 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.

[1867] 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.

[1868] 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.

[1869] 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.

[1870] FIG. 9 illustrates 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 behaviors 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.

[1871] 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.

[1872] 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).

[1873] 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.

[1874] 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."

[1875] 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.

[1876] 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).

[1877] 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.

[1878] 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.

[1879] Alternatively, the specific processing program 56 may be stored in a storage device such as a server connected to the data processing device 12 via the network 54, and the specific processing program 56 may be downloaded and installed on the computer 22 in response to a request from the data processing device 12.

[1880] It is not necessary to store all of the specific processing program 56 in a storage device such as a server connected to the data processing device 12 via the network 54, or to store all of the specific processing program 56 in the storage 32; only a portion of the specific processing program 56 may be stored.

[1881] The hardware resource for executing a specific process can be any of the following processors: An example of a processor is a CPU, which is a general-purpose processor that functions as a hardware resource for executing a specific process by executing software, i.e., a program. Another example of a processor is a dedicated electrical circuit, such as an FPGA (Field-Programmable Gate Array), a PLD (Programmable Logic Device), or an ASIC (Application Specific Integrated Circuit), which is a processor with a circuit configuration designed specifically for executing a specific process. Each processor has built-in or connected memory, and each processor uses the memory to execute the specific process.

[1882] The hardware resource that executes the specific processing may be configured with one of these various processors, or may be configured with a combination of two or more processors of the same or different types (for example, a combination of multiple FPGAs, or a combination of a CPU and an FPGA). Also, the hardware resource that executes the specific processing may be a single processor.

[1883] As an example of a system configured with a single processor, first, one processor is configured by combining one or more CPUs and software, and this processor functions as a hardware resource that executes a specific process. Second, there is a system that uses a processor that realizes the functions of an entire system including multiple hardware resources that execute a specific process on a single IC chip, as typified by SoC (System-on-a-chip). In this way, a specific process is realized using one or more of the above-mentioned various processors as hardware resources.

[1884] Furthermore, the hardware structure of these various processors can be, more specifically, an electric circuit that combines circuit elements such as semiconductor devices. The specific processing described above is merely an example. Therefore, it goes without saying that unnecessary steps may be deleted, new steps may be added, or the processing order may be rearranged, without departing from the spirit of the invention.

[1885] The above-described description and illustrations are a detailed explanation of the parts related to the technology of the present disclosure and are merely an example of the technology of the present disclosure. For example, the above description of the configuration, functions, actions, and effects is an explanation of an example of the configuration, functions, actions, and effects of the parts related to the technology of the present disclosure. Therefore, it goes without saying that unnecessary parts may be deleted, new elements may be added, or replacements may be made to the above-described description and illustrations within the scope of the gist of the technology of the present disclosure. Furthermore, to avoid confusion and facilitate understanding of the parts related to the technology of the present disclosure, the above-described description and illustrations omit explanations of common technical knowledge that do not require particular explanation to enable the implementation of the technology of the present disclosure.

[1886] All publications, patent applications, and technical standards mentioned in this specification are herein incorporated by reference to the same extent as if each individual publication, patent application, or technical standard was specifically and individually indicated to be incorporated by reference.

[1887] The following is further disclosed regarding the above embodiment.

[1888] (Claim 1)

[1889] means for receiving email messages and social networking service messages;

[1890] A method for analyzing received messages to detect potentially fraudulent wording and unreliable URLs, and

[1891] a way to alert you to potential fraud;

[1892] A means to respond to calls from anonymous or unknown phone numbers with an automated voice and analyze the content of the call in real time;

[1893] A system that includes a means to learn from past fraud cases and update the model to respond to new fraud techniques.

[1894] (Claim 2)

[1895] The system of claim 1 further comprising means for initially responding to calls from withheld or unknown phone numbers with an automated voice message and requesting confirmation of the call in order to notify the user of the contents of the call if there is a possibility of fraud.

[1896] (Claim 3)

[1897] 10. The system of claim 1, further comprising a natural language processing algorithm for assessing the likelihood of fraud based on the analysis of received messages and calls.

[1898] "Example 1"

[1899] (Claim 1)

[1900] means for receiving email messages and social networking service messages;

[1901] A method for analyzing received messages to detect potentially fraudulent wording and unreliable URLs, and

[1902] a way to alert you to potential fraud;

[1903] A means to respond to calls from anonymous or unknown phone numbers with an automated voice and analyze the content of the call in real time;

[1904] A means to learn from past fraud cases and update the model to respond to new fraud methods,

[1905] a means for displaying a warning to a user in the event of a possible fraud;

[1906] A means of analyzing messages and calls in real time using natural language processing algorithms to detect potential fraud;

[1907] A means of detecting new fraud methods using generative AI models based on the analysis results; and

[1908] A system including:

[1909] (Claim 2)

[1910] The system of claim 1 further comprising means for initially responding to calls from withheld or unknown phone numbers with an automated voice message and requesting confirmation of the call in order to notify the user of the contents of the call if there is a possibility of fraud.

[1911] (Claim 3)

[1912] 10. The system of claim 1, further comprising a natural language processing algorithm for assessing the likelihood of fraud based on the analysis of received messages and calls.

[1913] "Application Example 1"

[1914] (Claim 1)

[1915] means for receiving email messages and social networking service messages;

[1916] A method for analyzing received messages to detect potentially fraudulent wording and unreliable URLs, and

[1917] a way to alert you to potential fraud;

[1918] A means to respond to calls from anonymous or unknown phone numbers with an automated voice and analyze the content of the call in real time;

[1919] A means to learn from past fraud cases and update the model to respond to new fraud methods,

[1920] A system including means for visually alerting a user to a risk of fraud using a user-worn display device.

[1921] (Claim 2)

[1922] The system of claim 1 further comprising means for initially responding to calls from withheld or unknown phone numbers with an automated voice message and requesting confirmation of the call in order to notify the user of the contents of the call if there is a possibility of fraud.

[1923] (Claim 3)

[1924] 10. The system of claim 1, further comprising a natural language processing algorithm for assessing the likelihood of fraud based on the analysis of received messages and calls.

[1925] "Example 2: Combining Emotion Engines"

[1926] (Claim 1)

[1927] means for receiving email messages and social networking service messages;

[1928] A method for analyzing received messages to detect potentially fraudulent wording and unreliable URLs, and

[1929] a way to alert you to potential fraud;

[1930] A means to respond to calls from anonymous or unknown phone numbers with an automated voice and analyze the content of the call in real time;

[1931] A means to learn from past fraud cases and update the model to respond to new fraud methods,

[1932] A means of analyzing the user's emotional state and adjusting the intensity and timing of alerts;

[1933] The system includes a means for generating a warning message for a user based on the results of analysis of the call content and messages.

[1934] (Claim 2)

[1935] The system of claim 1 includes a means for first res...

Claims

1. means for receiving email messages and social networking service messages; A method for analyzing received messages to detect potentially fraudulent wording and unreliable URLs, and a way to alert you to potential fraud; A means to respond to calls from anonymous or unknown phone numbers with an automated voice and analyze the content of the call in real time; A system that includes a means to learn from past fraud cases and update the model to respond to new fraud techniques.

2. The system of claim 1 further comprising means for initially responding to an incoming call from an unnamed or unknown telephone number with an automated voice message and requesting confirmation of the call in order to notify the user of the contents of the call if there is a possibility of fraud.

3. 10. The system of claim 1, further comprising a natural language processing algorithm for assessing the likelihood of fraud based on an analysis of received messages and calls.

Images