Fraud prevention device

Abstract

To obtain a fraud prevention device capable of preventing special fraud without applying cost or burden to a user such as an aged person.SOLUTION: A fraud prevention system 1 detects fraud information so as to send a notification to a third person. The system includes: a camera 45 for storing a learning model 40B obtained by performing machine learning regarding a plurality of pieces of fraud information used in a fraud; information detection sections 41 for detecting peripheral detection information; a conversion section 43 for converting detection information from the information detection section 41 into recognizable information to be recognized; an artificial intelligence section 44 for inputting the recognizable information to the learning model 40B so as to acquire and determine answer information showing whether the output information is fraud information or not; and a communication section for reading registration information on a notification destination of the third person when the answer information is fraud information, so as to send the answer information to a registration destination of the registration information via a predetermined communication network.SELECTED DRAWING: Figure 3

Description

[Technical field]

[0001] The present invention relates to fraud prevention devices. [Background technology]

[0002] In recent years, there have been many cases of specialized fraud in which people pretend to be sons or grandsons and demand money, and the amount of damage caused has become enormous.

[0003] For example, on the first phone call, they might say something like, "I have a cold and my throat is sore," so that their different voice doesn't seem unnatural.

[0004] There are devices that detect such special frauds. For example, Patent Document 1 describes a suspicious call discrimination system that acquires the contents of a call from the other party, and if the other party is from a wrong affiliation and the contents of the call contain a prohibited keyword, displays notification information on the user's phone indicating that the call is suspicious.

[0005] Specifically, the system includes a telephone device connected to a computer device, a router for connecting the computer device and the telephone device to a network such as the Internet, and a database server connected to the network.

[0006] When an incoming call is detected, the computer device starts up and acquires the contents of the call, and extracts from the contents of the call affiliation information relating to the organization to which the other party of the call claims to belong.

[0007] Next, an organization database in which organization information relating to organizations is registered is referenced to determine whether the affiliation information includes any erroneous information.

[0008] If it is determined that the affiliation information does not include any erroneous information, a NG keyword database is referenced in which NG keyword information is registered regarding NG keywords representing words used in calls from individuals who misrepresent their affiliation to an organization.

[0009] Next, it is determined whether the call contents contain prohibited keywords, and if it is determined that the affiliation information contains incorrect information or if it is determined that the call contents contain prohibited keywords, it is notified that the call is suspicious and a message to that effect is displayed on the display unit. [Prior art documents] [Patent documents]

[0010] [Patent Document 1] JP 2007-228384 A Summary of the Invention [Problem to be solved by the invention]

[0011] However, the system in Patent Document 1 simply displays on the telephone that the call is suspicious, and the elderly and others do not know what to do. In addition, the system in Patent Document 1 uses a large-scale computer device, which is costly.

[0012] The present invention has been made in consideration of the above problems, and aims to provide a fraud prevention device that is capable of preventing special frauds before they occur without imposing a burden on elderly users, etc., and without incurring costs. [Means for solving the problem]

[0013] The present invention has been made in consideration of the above problems, and the invention of claim 1 is a fraud prevention device that detects fraudulent information and notifies a third party, and is characterized in that it has a memory unit that stores a learning model that has been machine-learned to learn multiple fraudulent information used in frauds, an information detection unit that detects surrounding detected information, a conversion unit that converts the detected information from the information detection unit into recognizable information that can be recognized, an artificial intelligence unit that inputs the recognizable information into the learning model and obtains and determines whether the output information is the fraudulent information, and if the answer information is fraudulent information, a communication unit that reads the registered information of the third party's reporting destination and communicates the answer information to the registration destination of the registered information via a specified communication network.

[0014] The invention of claim 2 is a fraud prevention device that is connected to an LED and a speaker, and has an alarm unit that turns on the LED, outputs a sound from the speaker, or vibrates when the artificial intelligence unit determines that the information is fraudulent.

[0015] In the invention of claim 3, the artificial intelligence unit executes a risk determination process for determining the level of risk of the fraudulent information depending on the risk level, and a process for changing the output amount of the LED, speaker, or vibration intensity depending on the risk level.

[0016] In the invention according to claim 4, the predetermined communication network is an LTE (Long Term Evolution) line.

[0017] The invention according to claim 5 notifies a communication terminal such as a smartphone or a mobile phone via the LTE line via SNS (Social Networking Service), email, or SMS (Short Message). Effect of the Invention

[0018] As described above, according to the present invention, it is possible to prevent specialized fraud without the involvement of the parties involved in the fraud. [Brief description of the drawings]

[0019] [Figure 1] 1 is an external view of a fraud prevention device according to an embodiment of the present invention; [Diagram 2] 1 is a schematic configuration diagram of a fraud prevention system according to an embodiment of the present invention. [Diagram 3] 1 is a block diagram of a fraud prevention device according to an embodiment of the present invention. [Figure 4] FIG. 2 is a diagram illustrating machine learning of a learning model according to the present embodiment. [Diagram 5] A sequence diagram of the fraud prevention device of this embodiment. [Figure 6] FIG. 11 is a diagram of a message screen according to the present embodiment. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0020] Each embodiment will be described in detail below with reference to the accompanying drawings. In this embodiment, the description of already known technologies will be omitted. In addition, the following are examples of devices and methods for embodying the technical idea of ​​the invention, and the technical idea of ​​the present invention is not limited to the following. The technical idea of ​​the present invention can be modified in various ways within the scope of the matters described in the claims. In particular, it should be noted that the drawings are schematic and may differ from the actual product.

[0021] Various embodiments of the present invention will now be described with reference to the drawings.

[0022] FIG. 1 is an external view of the fraud prevention device 40 of this embodiment. The fraud prevention device 40 is very simple and equipped with advanced technology. The fraud prevention device 40 is a computer, and is configured with a CPU (Central Processing Unit), a ROM (Read-only Memory) that stores control programs and the like that run on the CPU, and a RAM (not shown) for temporarily storing various data. It also includes a main body and a power cord connected to the main body. The device is always powered on and analyzes data from input sensors (voice, image, temperature, smell) and contacts third parties. It uses an LTE (Long Term Evolution) line to directly contact relevant parties via email, SNS (Social Networking Service) (LINE (registered trademark), etc.), SMS (short message), etc. The inside of the main body may have the same configuration as a smartphone or mobile phone. That is, when the device is powered on and it is determined that there is fraud, the destination information (telephone number, email address, etc.) is automatically entered, and it is possible to automatically transmit fraud information to the third party's smartphone or mobile phone. Furthermore, the main body is provided with an artificial intelligence unit 44 (see FIG. 3, etc.), which is configured to be able to determine, for example, based on a conversation on the telephone set T, whether or not the conversation is related to fraud.

[0023] As shown in Fig. 1, it is preferable to install the fraud prevention device 40 in the living room (where the elderly person or the like usually stays). Fig. 1 shows an example in which it is installed on a telephone table (not shown) in the living room. Furthermore, the fraud prevention device 40 is configured to determine whether fraud is occurring based on voice, body temperature, smell, and facial expression (image) data from everyday conversations in the user's living room or entrance.

[0024] As shown in Figure 2, the fraud prevention system 1 includes a fraud prevention device 40 (tabletop type, etc.). This fraud prevention device 40 is always ready to start by connecting a USB power adapter to an AC outlet (not shown). Note that the full power supply is turned on only when it is determined that fraud has occurred, and the device may be configured to normally determine whether or not fraud has occurred using a low current. This is because electricity costs are low.

[0025] Further, the fraud prevention device 40 is connected to an information detection unit 41, a camera 45, a microphone 42, an odor sensor 41A, a temperature and humidity sensor 41B, and the like.

[0026] The fraud prevention device 40 can be replaced by a device incorporating a camera 45, microphone 42, speaker 46, vibration unit 46A, etc. (for example, a smartphone or other highly functional mobile terminal with a dedicated app downloaded is also within the technical scope of the present invention) as long as it can fully function as a microphone and camera. In this embodiment, an example is described in which a camera 45, microphone 42, speaker 46, temperature and humidity sensor 41B, odor sensor 41A, etc. are connected. Note that the camera 45, microphone 42, speaker 46, temperature and humidity sensor 41B, etc. may be provided outside the main body, inside, or outside.

[0027] The fraud prevention device 40 is also configured to execute SMS, email or SNS (Social Networking Service) communication 20 via the LTE line 50, and to be able to communicate with the third party mobile terminal 10. Here, in this specification, "SNS" refers to social network services in general that have email chat functions and voice call functions and are commonly used both domestically and internationally, such as LINE (registered trademark) and WeChat (registered trademark).

[0028] Fig. 3 is a block diagram of the fraud prevention device 40. As shown in Fig. 3, the overall management unit 40A of the fraud prevention device 40 includes a database that stores a learning model 40B, an artificial intelligence unit (abnormality diagnosis) 44, an abnormality communication unit 40D, an abnormality warning unit 40E, and the like.

[0029] The learning model 40B stores keywords such as money, accounts, credit cards, etc., so that it can correspond to a large amount of information. These diagnostic keywords are preferably maintained by, for example, the person himself or his child (or another person) by periodically operating the fraud prevention device 40. It is configured to be compatible with a new version by remotely updating the device, including updating the keywords, and restarting it.

[0030] The artificial intelligence unit 44 makes a comprehensive judgment after the detection information (audio, image, etc.) from the microphone 42, etc. (camera 45, odor sensor 41A, temperature / humidity sensor 41B, etc.) indicates an incoming call and the receiver of the telephone T is picked up, and then judges that a conversation has started. The unit may be configured to turn on the power when it detects that a conversation has started (or may be configured to normally operate at a minimum level in sleep mode, and to turn on the power fully upon detection). After that, the detection information from the microphone 42, etc. (camera 45, odor sensor 41A, temperature / humidity sensor 41B, etc.) is converted by the conversion unit 43 into recognizable information that can be recognized by a computer and input to the artificial intelligence unit 44, and if it is determined to be fraud, the abnormality communication unit 40D and the abnormality warning unit 40E are activated.

[0031] When the abnormality communication unit 40D is activated by the artificial intelligence unit 44, it transmits information to the effect that an abnormality has occurred to the mobile terminal 10 of a third party.

[0032] When the abnormality warning unit 40E is activated by the artificial intelligence unit 44, it activates the vibration unit (vibration motor) 46A to vibrate, generates a warning sound from the speaker 46, and lights up the LED or display unit (collectively referred to as the light-emitting unit).

[0033] As shown in Figure 4, in machine learning, detection information (voice, image data, etc.) from a microphone 42 or the like (camera 45, odor sensor 41A, temperature / humidity sensor 41B, etc.) is used as input information (voice, image data, etc.) 71, which is input to an input unit 72, and fraud data 74 indicating that a fraud has actually occurred is output by an output unit 73 as teacher data for learning by a learning model 40B. This learning model 40B may perform machine learning by collecting fraud information of all users on a single server or the like. The information is then periodically downloaded to the fraud prevention device 40 via an external communication unit 60.

[0034] The following will be explained using the sequence diagram in Fig. 5. The fraud prevention device 40 is activated by connecting the USB power adapter to an AC outlet (not shown). Although there are various types of detection information (audio, image, etc.) from the microphone 42 and the like (camera 45, odor sensor 41A, temperature / humidity sensor 41B, etc.), the following explanation will focus on audio as a representative example.

[0035] The artificial intelligence unit 44 acquires various signals such as voice signals and images from various sensors such as the microphone 42 and the camera 45, and the conversion unit 43 converts them into text data. In this state, when the telephone T receives an incoming call and the user picks up the receiver and speaks (d10), the artificial intelligence unit 44 inputs the text data of the voice data and various data into the learning model 40B for diagnosis (d12). Here, the voice data is converted directly into text data, and when it is determined that the image of the camera 45 is identical or similar to an image template with which pre-registered text data is associated, the text data associated with the image is input into the learning model 40B for diagnosis. In the case of the odor sensor 41A and the temperature / humidity sensor 41B, when the sensor is identical or similar to the registered template, the associated text data is input into the learning model 40B for diagnosis.

[0036] Then, based on the stored output information (answer information) of the learning model 40B, it is diagnosed whether the conversation is fraudulent or not (d14). For example, if the conversation content includes keywords such as money, transfer, and card, the learning model 40B judges it to be fraud.

[0037] If fraud is detected in the diagnostic step d14, the abnormality notification unit 40D and the abnormality warning unit 40E are activated, and when the abnormality warning unit 40E is activated by the artificial intelligence unit 44, it activates the vibration unit (motor) 46A to vibrate, generates a warning sound from the speaker 46, and turns on the LED or display unit 46C (collectively referred to as the light generating unit) (d16). At the same time, it communicates text and voice to the outside via communication (SNS, email, etc.). On the other hand, if fraud is not detected from the content of the conversation, etc., the abnormality notification unit 40D and the movement warning unit 40E are not activated, and the fraud prevention device 40 goes into standby mode.

[0038] Then, the abnormality notification unit 40D determines the destination of the abnormality notification and determines to notify the relevant person, who is the destination, by SNS (email, etc.) using the mobile phone line network of the machine. In other words, it determines that the abnormality notification is to notify the previously set contact point that an abnormality has occurred (d18).

[0039] Then, the report is sent to the determined report destination, for example, via the SNS communication 20 (see FIG. 2) (d20, d22). At this time, the contents of the conversation may be sent by voice or email.

[0040] This will cause the third party (child, etc.) to immediately contact the parent or relevant party, so the parent will immediately know that the previous phone call was a scam, thus preventing special frauds from occurring.

[0041] In this embodiment, since odor sensor 41A and temperature / humidity sensor 41B are provided, it is possible to know whether the parent's body temperature is rising due to agitation.

[0042] It can also tell if someone is stewing or grilling something in the kitchen and has left the gas on.

[0043] Furthermore, the artificial intelligence unit 44 determines the abnormality level from the contents of the conversation (for example, level 1, level 2, level 3).

[0044] For example, the degree of danger for a conversation is determined based on the number of specific keywords contained in the conversation, and a danger response process is executed according to the determined degree of danger.

[0045] Weighting is applied to the number of specific keywords contained in the conversation, and it responds to the number of fraud-related keywords such as "money" and "transfer." In other words, if the number is between 5 and 20, the conversation is judged to have a risk level of "1," and an email sending process appropriate for a risk level of "1" is executed.

[0046] In the SNS (email sending, etc.) process, an email is sent to the mobile device of a third party (child, etc.) indicating that a special fraud may currently be occurring. Figure 6 shows a screen displaying the sent message, etc. (showing the case of a normal conversation). If there is a risk of fraud, messages such as "The conversation has ended," "The tone of voice was excited," "The call duration was longer than usual," the starting word "telephone," the detected word "money," etc. are displayed on the screen of the mobile device 10 of the third party.

[0047] The more specific words and phrases are included in the conversation, the higher the risk is judged to be, and the higher the risk level is expressed as a numerical value. If the numerical value is 20 or more, the risk level of the conversation is judged to be "2" and an abnormality warning corresponding to the risk level of "2" is issued, for example by emitting an alarm sound.

[0048] In other words, various processes for preventing fraud are automatically executed based on the number of specific keywords used in specialized frauds contained in the conversation.

[0049] In this way, it is possible to prevent specialized fraud without involving the parties to the fraud, such as users.

[0050] The above explanation of Figs. 1 to 5 is merely for the purpose of deepening understanding of the contents of the present invention. Specifically, the present invention is embodied in each of the embodiments described below, and may be embodied in various modified forms without substantially departing from the principles of the present invention. All such modified forms are included in the scope of the present invention and the disclosure of this specification. The notification destination may be a public institution. [Explanation of symbols]

[0051] 1. Fraud Prevention System 10. Third Party Mobile Devices 20 Social Media Communication 30 Telephone exchange sites 40 Anti-Fraud Device 50 LTE lines T phone

Claims

1. A fraud prevention device that detects fraud information and notifies a third party, A memory unit that stores a learning model that is machine-learned from a plurality of pieces of fraud information used in fraud; An information detection unit that detects surrounding detection information; a conversion unit that converts the detected information from the information detection unit into recognizable information; An artificial intelligence unit that inputs the recognizable information into the learning model and obtains and judges whether the output information is the fraudulent information; The fraud prevention device has a communication unit which, if the answer information is fraud information, reads the registration information of the third party's reporting destination and communicates the answer information to the registration destination of the registration information via a specified communication network.

2. The fraud prevention device as described in claim 1, further comprising an alarm unit which is connected to an LED and a speaker and which turns on the LED, outputs a sound from the speaker, or vibrates when the artificial intelligence unit determines that the information is fraudulent.

3. The fraud prevention device of claim 1, wherein the artificial intelligence unit executes a risk determination process for determining a level of risk of the fraud information according to the risk level, and a process for changing the output amount of an LED, a speaker, or vibration strength according to the risk level.

4. 2. The fraud prevention device according to claim 1, wherein the predetermined communication network is an LTE (Long Term Evolution) line.

5. The fraud prevention device according to claim 4 , wherein the fraud prevention device notifies a mobile phone via SNS (Social Networking Service) or SMS over the LTE line.

Images