Usage history management and preservation method in ai system

Abstract

To provide a method for protecting the privacy of ordinary citizens in an artificial intelligence system which receives input of big data constructed from data containing personal privacy such as security camera footage, credit card usage histories, and smartphone usage histories and analyzes such big data, and countermeasures against runaway (singularity) of the artificial intelligence system.SOLUTION: There is provided a mechanism in which an artificial intelligence system which receives input of data containing privacy information is required to obtain authorization from a recording device before outputting information such as alerts or answers to questions, and in which the entire sequence of processing is recorded in a tamper-proof and verifiable manner, thereby making the AI system's behavior (information output) verifiable. It is also possible to early detect signs of AI runaway behavior and gain time to take countermeasures against the AI runaway behavior by monitoring the output of the artificial intelligence system regardless of whether the input data contains privacy information or not.SELECTED DRAWING: None

Description

[Technical Field]

[0001] This is a technical field related to privacy protection in artificial intelligence systems that receive input data including privacy information. The present invention relates to the technical field of privacy protection in artificial intelligence systems that receive input data that includes privacy information. The present invention provides a mechanism that requires an artificial intelligence system that receives input data that includes privacy information to obtain permission for information output from a recording device when outputting information (alerts, answers to questions, etc.), and further provides a mechanism that makes it possible to reveal any fraudulent behavior of the artificial intelligence (information output) by recording the series of processes. [Background technology]

[0002] In the following patents, the invention for "complete storage of browsing history" discloses an invention for a "method for storing browsing history" that targets street network cameras and aims to protect the privacy of ordinary citizens. The present invention expands on this by providing a method for protecting the privacy of ordinary citizens in an artificial intelligence system that receives input of big data constructed from data containing privacy, such as security camera images, credit card usage history, smartphone usage history, and location information, and analyzes the big data.

[0003] [Patent Document 1] Patent No. 6671627, Patent Application No. 2015-167298 (EJK-008), "Camera system enabling privacy protection", Applicant: NPO e-Cybersecurity Network Research Association

[0004] Even in today's technologically advanced world, children continue to go missing. Children often go missing in public spaces, such as the street (on their way to school or near their home), parks, event venues, and shopping malls. Even if a child goes missing inside a home, they cannot be abducted or moved to another location unless they pass through public spaces, such as roads around the home.

[0005] A "camera-monitored public space" is defined as a public space where network cameras are densely installed so that every person and vehicle in the area is photographed and recorded by one or more network cameras.

[0006] In 2005, the inventors proposed building a society in which it would be impossible for abductions and other crimes to go unreported, by having each citizen take pictures of the area around their home with a camera connected to their PC and store several weeks' worth of images on the PC. Furthermore, in 2010, with the aim of protecting the privacy of ordinary citizens, they developed the "e-Vigilant Camera," a standalone camera that does not require a PC and allows for the separation of the "camera administrator (assuming a city hall)" and the "image viewer (assuming a police station)" by encrypting and storing captured images. Furthermore, in 2016, they proposed the concept of "complete recording of browsing history" as a powerful measure to prevent abuse by administrators.

[0007] With current scientific technology, it is believed that "capturing and recording every movement in public spaces with cameras" is relatively easy and low-cost to achieve. In 2019, it was predicted that the street camera installation situation by the end of 2021 would be as follows: Tokyo would have approximately 1 camera per 1,000 people, London would have approximately 13 cameras per 1,000 people, and China as a whole would have approximately 439 cameras per 1,000 people. China is experiencing exceptionally high installation densities. Even in democratic countries such as Tokyo and London, the situation is far below the national average for China.

[0008] In China, it seems as though "camera-monitored public spaces" are on the verge of being achieved without any protection of the privacy of ordinary citizens. Around the world, except in China, the density of street cameras installed is far from the level required to "capture and record every movement in public spaces." Even so, the risks of privacy violations are being pointed out, and measures to protect privacy are being considered. The introduction of street cameras with facial recognition functionality is also progressing, and the introduction of regulations to protect privacy is also being considered.

[0009] The positive effects of "filming and recording all movements in public spaces with cameras" include improving social safety and efficiency. However, at the same time, for democratic nations, the negative effects include enabling system administrators and operators to "operate and use systems in a way that violates the privacy of ordinary citizens" and raising concerns about the emergence of a surveillance society.

[0010] First, we discuss the most basic utilization of "camera-monitored public spaces," namely, "the continuous tracking of criminal suspects and missing persons by on-demand image transfer from each camera."

[0011] In this case, the basic specifications for a "network camera system for monitoring public areas" that can be readily implemented are as follows: (1) Images taken by each camera are stored for a certain period of time. (2) The oldest image files are overwritten by newer image files. (3) Each camera will transmit recorded images via the Internet upon request from the administrator. Most of the images show "ordinary street scenes" and are expected to be deleted without ever being used. The reason for transferring images on demand is to reduce the amount of data traffic required to transfer recorded image files.

[0012] Figure 1 shows an example of a "camera-monitored public area" without privacy protection features. The "camera-monitored public area" is intended to use on-demand image transmission from each camera to track criminal suspects and missing persons.

[0013] In this example, we assume that people and vehicles are tracked using "relay-based tracking" by city hall employees and police station employees authorized by the city hall. Japan is divided into 47 prefectures, each of which is further divided into approximately 1,700 cities, towns, and villages. Streetlights are generally installed and operated by cities, towns, and villages, and each municipality typically has a police station. In this example, we assume that the streetlights installed and operated by city halls are replaced with "streetlight units," streetlights equipped with network cameras. We assume an area where streetlights are installed densely and the illuminated areas overlap. This applies to many urban areas and residential complexes in Japan. The "streetlights equipped with network cameras," installed, managed, and operated by each city hall, can be accessed via a "viewing terminal" installed in the city hall. This system consists of two components: [A] streetlight units, and [B] viewing terminals / viewing software.

[0014] [A] Street light unit The streetlight unit contains several CCD cameras that cover the area illuminated by the streetlights. Figure 2 shows a schematic diagram of the streetlight unit. Images captured in the streetlight unit over the past week to month are encrypted and overwritten. Images are transferred on demand. Image files are sent to the viewing terminal in response to a request from the terminal. It is expected that most image files will be overwritten and deleted without ever being viewed. This reduces the amount of image data transmitted and the burden on the network. By overlapping the streetlight illumination area with the camera's field of view, lighting for camera photography at night can be eliminated. By waiting until the periodic streetlight replacement period, the camera installation costs can be reduced. When the fields of view of each camera overlap, it is believed that it is possible to "track the shadows of people and vehicles" even with images captured by low-cost, low-resolution cameras. High-resolution cameras installed in key locations can also easily identify people's faces and vehicle license plates.

[0015] [B] Viewing device / viewing software The viewing terminal / viewing software (Figure 3) is operated by the viewer (such as a city hall employee or a police station employee authorized by city hall). By issuing an image viewing request, "images" are received from streetlights. Cameras are switched by the viewer clicking on the camera (position and viewing direction) displayed on the interactive map. To ensure smooth tracking, when tracking forward in time, the image taken by the displayed "next camera" will be taken a few seconds before the image displayed by the "previously selected camera."

[0016] The following three points can be listed as the main characteristics, advantages, and disadvantages of operating the "public area monitored by cameras" shown in Figure 1 using the on-demand image transfer method.

[0017] [1] It allows for chain tracking Since people and vehicles that leave the field of view of one camera continuously enter the field of view of the next camera, it is believed that smooth tracking will be possible by switching cameras in a relay fashion. In other words, it is believed that it will be possible to track any person or vehicle in a chain-like manner. As long as recorded images remain, it will be possible to track any person or vehicle by switching cameras and moving forward or backward in time, starting from any time.

[0018] [2] Improving social safety It will be possible to track a missing child based on the "location and time immediately before they went missing," pinpoint their current location, and quickly rescue them. Furthermore, street criminals such as snatchers, molesters, and graffiti perpetrators will almost always be identified and arrested. If this becomes widely known, those who do not want to be arrested will be discouraged from committing crimes, and as a result, the number of crimes is expected to drop dramatically. In this way, the realization of "public spaces monitored by cameras" is expected to greatly improve "social safety."

[0019] [3] Risk of privacy violation The ability to track every person and vehicle also means that the privacy of ordinary citizens will be seriously endangered. Authoritarian countries that do not place importance on personal privacy will likely move forward without hesitation with the realization of "camera-monitored public spaces." Democratic countries, however, will likely hesitate to implement "camera-monitored public spaces" due to concerns about privacy violations, potentially delaying improvements in social safety.

[0020] For a "camera-monitored public space" to be accepted by the public and realized in a democratic nation, the following conditions are likely to be necessary. [Condition-A] Surveillance of civilians and passersby is permitted only for predetermined purposes. [Condition-B] System usages by the administrators and operators can be tracked by the people, and the misuse of the system is exposed.

[0021] To satisfy the above [Condition-a] and [Condition-b], a system that protects the privacy of ordinary citizens must be reliable. To achieve this, the following is considered necessary:

[0022] (1) Social discussion will define the purpose and scope of permissible use of each "public space monitored by cameras." Any other use will be considered misuse and prohibited. Uses that contribute to "social safety," such as tracking criminals or missing persons, will be considered to be the purpose and scope of permissible use. On the other hand, tracking ordinary citizens who have not been officially recognized as criminals or missing persons will be considered misuse and will be subject to severe penalties. It will be essential that "judgments" on each individual case be made in accordance with the law by a court or a similar newly established government agency.

[0023] (2) "Network camera systems for monitoring public spaces with anti-abuse functions" will be developed and maintained by private companies under government approval based on law. From the perspective of reducing costs and preventing monopolies, it is preferable for multiple businesses to participate.

[0024] (3) The operation of each authorization system will be fully standardized, and components will be fully certified and registered. (4) In order to prevent the theft and use of image files by third parties such as hackers, encryption of image files and communications is used, just as with other valuable information exchanged over the Internet.

[0025] (5) In order to prevent abuse by system owners and operators, "system operations, including image viewing history," will be recorded in a verifiable and tamper-proof form, such as using blockchain technology. (6) The "records of system operation, including image viewing history" will be made available for public and citizen verification. It would be ideal to have an independent body dedicated to conducting reliable and rigorous verification.

[0026] Figure 4 shows the "Network camera system for monitoring public spaces" shown in Figure 1. [Patent Document 1] This figure shows a schematic diagram of a system that implements a "complete recording of browsing history" function for privacy protection based on the above. The "network camera system for public space surveillance with privacy protection" consists of three components: [A] the street light unit, [B] the viewing terminal / viewing software, and [C] a recording device, as shown in Figure 1. The three components (devices) operate in accordance with the system's own operating standards and are connected to each other over the Internet using the system's own encrypted communication standards. In order for each device to be connected to the system, it must be pre-inspected to ensure compliance with the system standards and pre-registered in the system. When the camera owner or administrator attempts to view images recorded on the camera using a viewing device, the viewing activity is recorded in a verifiable and tamper-resistant form on a recording device operated by an independent "recorder," and is made available for verification by society and citizens. In particular, using a recording method using blockchain is expected to make the record tamper-proof.

[0027] In this example, it is assumed that tracking of people and vehicles using "chain tracking" will be carried out by city hall employees and police station employees authorized by city hall. Each of the three components of this system, the "street light unit," "viewing terminal / software," and "recording server," must pass a preliminary inspection in order to be registered in the system. They operate in accordance with the system's own operating regulations and are connected to each other via the Internet based on the system's own encryption communication regulations.

[0028] [A] Street light unit Although the external appearance is the same as the street light unit shown in Figures 1 and 2 above, the operating algorithm implements the "complete recording of browsing history" function for privacy protection shown in Figure 4. When a request for image transmission is made from a viewing terminal, the image file is sent to the viewing terminal under the conditions described in the attached "authorization code."

[0029] [B] Viewing device / viewing software While the external appearance is the same as the street light units shown in Figures 1 and 3 above, the operating algorithm implements the "complete recording of browsing history for privacy protection" function shown in Figure 4. When an image viewing request is input, the viewing terminal sends it to the recording server as a "Request code," and receives an "authorization code" from the recording server. The viewing terminal then forwards the "authorization code" to the street light, and receives an "encrypted image" from the street light. The viewing terminal decrypts the "encrypted image" using an encryption key and displays it on the screen. It is also possible to control the clarity of the output image depending on the level of the viewer. For example, one possible operation would be to display a pixelated image for maintenance workers, allowing them to see the camera's operation but preventing them from seeing the details of the people and vehicles in the image.

[0030] [C] Recording Server When the "recording server" receives a "request code" from the viewing terminal, it sends an "authorization code" to the viewing terminal in accordance with the "operational regulations" prescribed in advance by the city hall. The "operational regulations" describe the "conditions for utilizing image files" established by the city hall, and the "recording server" automatically follows these regulations. The "recording server" stores the entire process of receiving the "request code," making a judgment based on the "operational regulations," and sending the "authorization code" as a "viewing history" for each viewer ID and image file ID. In particular, blockchain storage

[15] enables highly reliable "viewing history storage." In this case, to prevent the blockchain from becoming too large and its operating costs from rising, it is also effective to update the blockchain to a new one periodically (e.g., monthly). Furthermore, the costs associated with the blockchain recording system could be covered by city halls, which are the requesting parties, paying each time a request is made.

[0031] There only needs to be one "recording server" in place nationwide. The "recording server" will never touch the "encrypted image files" owned by each city hall, nor will it know the encryption keys set by each city hall. "Browsing history" will be made available for audits by auditing agencies and civic groups, and they will check whether information obtained in "camera-monitored public spaces" managed and operated by the city hall in question has been misused. Any misuse will be exposed, and those involved will be punished.

[0032] The following types of cameras are expected to be used in "camera-monitored public spaces." * Built-in street light * Freestanding type for indoor and outdoor installation * Car-mounted and drone-mounted * Built-in smartphone, wearable The following is an overview of the functions of the viewing system for images taken and saved by the various cameras (fixed and mobile) mentioned above. * The location and time of the camera that captured the recorded image displayed in the window, as well as the locations of nearby cameras that capture images around that time, are displayed on an interactive map. * Images taken at any location and time can be viewed by smoothly changing the location (camera) and time.

[0033] If each city hall and police station were to individually manage and operate "camera-monitored public spaces," then having a specialized contract agency that could track people and vehicles 24 hours a day would enable rapid tracking of people and vehicles over a wide area, which would be effective in responding quickly to incidents and accidents. One such specialized agency per country would be sufficient, and it is expected that it would specialize in visual tracking using viewing devices and automatic tracking using software. In such cases, it would be necessary to create rules and mechanisms that would allow the specialized agency to smoothly use the image viewing rules and image encryption keys that each city hall sets independently.

[0034] It is quite possible that laws and rules may be enacted that require all cameras that view public spaces to comply with one of the "network camera systems for monitoring public spaces with anti-abuse features" and be pre-registered with the system, in order to protect the privacy of ordinary citizens. Cameras that operate in public spaces may be required to disclose information to passersby about the registration system, camera ID, manager / operator, purpose of operation, and operational track record, i.e., to introduce themselves.

[0035] Figure 5 is a diagram for explaining an embodiment not according to the present invention. In the system shown in Figure 5, the camera, viewing device / viewing software, and recording server are connected to the Internet and can communicate with each other. Information processing within the camera, viewing device / viewing software, and recording server follows a common standard so that common codes and files can be processed. It is important that this standard is kept secret. In particular, it is important that the encryption and decryption methods used by the camera, viewing device, and viewing software are kept secret.

[0036] The camera system, whose operation, processing, and communications are shown schematically in Figure 5, requires an authorization code issued by the recording server when sending, transferring, obtaining, or opening encrypted image files captured by a surveillance camera. In response to a request for an authorization code, the recording server issues an authorization code under predetermined authorization code issuance conditions. The authorization code issuance conditions can be set or changed by a predetermined person outside the recording server. The recording server records information from the reception of the request for an authorization code to the issuance of the authorization code, such as the camera ID, image file ID, viewer ID, authorization code issuance time, and the issued authorization code. Image processing within the surveillance camera analyzes the information contained in the image captured by the camera, and the analysis results are stored in association with the image and / or the analysis results are notified to the recording server.

[0037] The viewing condition setter can secretly set the authorization code issuance conditions on the recording server. The authorization code issuance conditions themselves are preferably encrypted. The encryption key setter can secretly set the encryption key on the camera. The encryption key setter also assigns the encryption key to the viewer. The encryption key itself is preferably encrypted. The viewing condition setter and the encryption key setter can be the same person; for example, the general affairs department or safety and security department of a city hall could be considered.

[0038] The viewer is assigned a unique viewer ID and an encryption key by the encryption key setter. The viewer decides whether or not to view the request from the recording server. If the viewer decides to accept the request, the viewing device and viewing software will carry out the processing and communication described below and output a blurred image, etc.

[0039] A unique camera ID is assigned to the camera. Images taken by the camera are encrypted using an encryption key. Each encrypted file is assigned a unique file ID that includes the camera ID and the time the image was taken. This file ID may be the same as the file name.

[0040] The captured images may be still images taken at a set interval, or videos recorded over a set period of time. The camera processes the captured images to detect suspicious activity such as fighting, arson, or loitering, and if any suspicious activity is detected, the camera requests predetermined viewers to view the images via the recording server.

[0041] In addition to the function of issuing a request to view images via communication over the Internet, a function of issuing an external alarm using a siren or rotating light installed near the camera, and / or a function of making emergency contact with the police or fire department via a mobile phone line may also be added. The image processing conditions, conditions for issuing a request to view, etc. are set in advance by the aforementioned viewing condition setter and encryption key setter.

[0042] Due to inaccuracies in the camera's clock, it is possible that two or more files may be generated at the same time. In order to ensure that a unique file ID is generated even in such cases, it is possible to add part of the information inside the captured image file, such as a random number, to the file ID. The encrypted file with the unique file ID is stored inside the camera.

[0043] The recording server processes the viewing request code input from the camera according to the pre-set authorization code issuance conditions and outputs the authorization code to the viewing device / viewing software. The authorization code is created by adding a viewing condition code that indicates the level of authorization to view the image to the viewer ID, file ID, and reason code built into the viewing request code. The level of blurriness of the viewable image is determined according to the level of viewing authorization specified by the authorization code.

[0044] The recording server records the process of interaction with the camera, viewing device, and viewing software. In particular, by recording the viewer ID, file ID, viewing condition code, and reason code, it records which viewer was permitted to view which image file, at what blur level, and for what reason. From an information security perspective, it is desirable that encrypted files containing the subject's private information and the encryption key set by the encryption key setter are never entered into the recording server. The recording server can be operated independently by a third party different from the viewing condition setter, encryption key setter, and viewer. Possible administrators of the recording server include private server providers, city hall information media departments, incorporated associations, and specified non-profit organizations.

[0045] The viewing device and viewing software can be a specially developed viewing device with viewing software installed, or viewing software installed on a commercially available laptop. In particular, using a specially developed viewing device with viewing software installed ensures even greater security. In response to a file viewing request from the recording server, the viewing device and viewing software communicate with the camera and recording server to obtain the encrypted file. A file conversion process is then performed on the encrypted file, generating an obscured file. The file conversion process is carried out in detail as follows:

[0046] First, the input encrypted file is decrypted using an encryption key, and the encryption is removed, creating a decrypted file in a common still image or video file format. Next, the decrypted file is further blurred using a method such as mosaic processing specified by the viewing condition code to create a blurred image. The blurred image generated by the file conversion process is output to the viewer. The blurring strength is specified by the viewing condition code. If the blurring strength is set to zero, a blurred image is output that is essentially unblurred. The blurred image is in a common still image or video file format. The viewing condition code is used separately from the authorization code input from the recording server. Note that in order to make it difficult for external parties to read the "decrypted file," which is a temporary file generated during the "file conversion process" of the viewing device and viewing software shown in Figure 5, it is effective to scramble the memory.

[0047] In this system, the information and communication process from the time the camera issues a request to view an image to the time the viewer receives the blurred image is as follows:

[0048] First, the camera processes the images it has taken and / or other sensing information such as temperature, vibration, sound waves, and electromagnetic waves according to a specified algorithm. If the results detect abnormalities such as fights, arson, loitering, molestation, burglary, or robbery, it will request that specified viewers view the images, etc.

[0049] The captured image in Figure 5 can be considered as sensing information such as captured images. Also, the image processing in Figure 5 can be considered as sensing information processing such as image processing.

[0050] The camera creates a viewing request code consisting of the viewer ID and file ID. It sends this viewing request code to the recording server via communication 2. Note that Figure 5 shows an example in which the viewing request code has only one file ID, but it may have multiple. Although not shown in the figure, the viewing request code also includes a reason code that describes the reason why the viewer should view the file. Possible reason codes include "abnormal movements that appear to be a fight have been detected" or "the camera's internal operation is strange and a malfunction is suspected."

[0051] The recording server that receives the viewing request code generates an authorization code through process 3 based on the predetermined authorization code issuance conditions and sends it to the viewing device / viewing software via communication 4. The authorization code is created by adding the reason code included in the viewing request code to the viewer ID, file ID, and viewing condition code that indicates the level of authority to view the image that are embedded in the inquiry code.

[0052] The viewing device / viewing software that receives the authorization code via communication 4 separates the reason code included in the viewing request code via process 4, displays it to the viewer, and asks them to decide whether or not to accept the viewing request from the camera. Alternatively, the viewing device / viewing software automatically decides whether or not to accept the viewing request in accordance with predetermined conditions. If the decision is made to accept the viewing request, the viewing device / viewing software transmits the authorization code via communication 5.

[0053] The camera receives the authorization code via communication 5 and, via process 5, selects the encrypted file corresponding to the file ID listed in the authorization code. The selected encrypted file is sent to the viewing device / viewing software via communication 6. The viewing device / viewing software receives the encrypted file via communication 6 and converts it into a blurred file via file conversion processing, before outputting it to the viewer. Mosaic processing is a suitable method for blurring images.

[0054] Summary of the Invention [Problem to be solved by the invention]

[0055] There are three possible ways to utilize "camera-monitored public spaces." [a] On-demand image transfer from each camera: Tracking criminal suspects and missing persons [b] Real-time image analysis by each camera: Detection of criminal activity, detection of wanted criminals by facial recognition [c] Data is transferred to the server, and big data analysis is performed using AI: detection of abnormal behavior, criminals, and terrorists.

[0056] In the above-mentioned "(a) On-demand image transfer from each camera," the "privacy-containing information to be output" is specified in advance, [Patent Document 1] A method based on the invention disclosed in makes it possible to protect the privacy of ordinary citizens.

[0057] However, in addition to street cameras, information from smartphones (location information, communication history, direct communication transactions), credit card usage history, and other types of information containing privacy concerns are being collected as big data and analyzed by artificial intelligence systems, and this trend is expected to accelerate further.

[0058] It is predicted that a "society in which every movement in public spaces is constantly monitored and sensed" will become technologically feasible through the use of AI utilizing IoT and big data. In a "society in which every movement in public spaces is constantly monitored and sensed," how to protect the privacy of ordinary citizens will become an issue that concerns the very foundation of democracy for democratic nations.

[0059] [Means for solving the problem]

[0060] To achieve the above objective, the invention related to claim 1 is an information output method characterized in that an information processing device / information processing software obtains permission for information output from a recording device / recording software before outputting information.

[0061] The invention related to claim 2 is the information output method described in claim 1, characterized in that the recording device and recording software have a firewall function that monitors, controls, and records the output.

[0062] The invention related to claim 3 is the information output method according to claim 1 or claim 2, characterized in that the recording device and recording software can be operated independently of the information processing device and information processing software.

[0063] The invention related to claim 4 is the information output method described in claims 1 to 3, characterized in that the recording device and recording software record a series of processes using a blockchain.

[0064] The invention related to claim 5 is the information output method according to any one of claims 1 to 4, characterized in that the information processing device / information processing software is an artificial intelligence system / artificial intelligence software that receives input of data including privacy information.

[0065] The invention related to claim 6 is the information output method described in any one of claims 1 to 4, characterized in that the information processing device and information processing software are a server system and server software that controls cameras.

[0066] [Effects of the Invention]

[0067] The purpose of this paper is to provide a method for protecting the privacy of ordinary citizens in artificial intelligence systems that receive data that includes the privacy of ordinary citizens.

[0068] In the coming hyper-information society, data containing private information about ordinary citizens will be collected and generated from sources such as credit cards, smartphones, self-driving cars, and street cameras and sensors. It is expected that this big data will be analyzed by artificial intelligence systems to produce a variety of outputs (such as estimating individuals' purchasing habits and detecting wanted criminals and terrorists). However, there is still no established method for protecting the privacy of ordinary citizens.

[0069] This invention provides a mechanism that requires an artificial intelligence system that receives input data including privacy information to obtain permission from a recording device before outputting information (such as an alert or answer to a question), and furthermore, by recording the entire process, makes it possible to reveal any irregularities in the behavior of the artificial intelligence (information output).

[0070] [Brief explanation of the drawings]

[0071] [Figure 1] Figure 1 is a schematic diagram of a "camera-monitored public area" that does not conform to the present invention and does not have a privacy protection function. [Figure 2] Figure 2 shows a schematic diagram of an LED street light unit with a built-in network camera. [Figure 3]Figure 3 is a schematic diagram of the operation screen of the viewing device and viewing software.

[0072] [Figure 4] FIG. 4 shows an example that is not based on the present invention, and is a "network camera system for monitoring public spaces" shown in FIG. 1. [Patent Document 1] This is a schematic diagram of a system that implements a "complete record of browsing history" function for privacy protection based on the above. [Figure 5] FIG. 5 shows an example not based on the present invention, and is similar to that shown in FIG. [Patent Document 1] This is an example of the operation and operation of a system that implements a "complete record of browsing history" function for privacy protection based on the above.

[0073] [Figure 6] Figure 6 shows Example 1, which is an example of claim 1. [Figure 7] Figure 7 shows Example 2, which is an example of claim 2. [Figure 8] Figure 8 shows Example 3, which is an example of claim 3. [Figure 9] Figure 9 shows Example 4, which is an example of claim 4.

[0074] [Figure 10] Figure 10 shows Example 5, which is an example of claim 5. [Figure 11] Figure 11 shows Example 6, which is an example of claim 6. [Figure 12] Figure 12 shows Example 7, which is an example of claims 1, 2, 3, 4, and 5.

[0075] DETAILED DESCRIPTION OF THE INVENTION [Example]

[0076] Example 1 shown in Figure 6 is an example of claim 1. Example 6 employs an information output method characterized in that an information processing device / information processing software obtains permission for information output from a recording device / recording software before outputting information obtained by calculating information from a street camera or smartphone.

[0077] As a method for storing utilization history for the "real-time image analysis by each camera" function, it is considered appropriate to store the "requested actions" of the "camera" as shown in Figure 6. The basic process for "saving utilization history" is as follows:

[0078] (1) The camera analyzes captured images in real time based on predefined rules and detects abnormalities. (2) If the camera detects an abnormality, it sends a request code to the recording server. (3) The "recording server" processes the "request code" based on pre-defined rules and sends an "authorization code" containing the permission to act to the "camera." (4) The "recording server" records the entire process and makes it available for public inspection. (5) The camera will take action (such as reporting or calling out) based on the content specified in the authorization code.

[0079] In Example 6, it is also possible to monitor children as a free government service at the request of their parents. For example, it is possible to keep track of a child's location and situation from the time they leave home until they return, and if anything unusual occurs, automatically call out to them, notify their parents, or call the police. For more reliable tracking, it is thought that it would be effective for each camera to communicate via Bluetooth with the child's smartphone. If the fields of view of each camera overlap, automatic tracking of people and vehicles is thought to be relatively easy, but it may also be possible to achieve automatic tracking even if there is no partial overlap. [Example]

[0080] Example 2 shown in Figure 7 is an example of claim 2. Example 2 employs the information output method described in claim 1, characterized in that the recording device and recording software have a firewall function that monitors, controls, and records output. Figure 7 shows an example in which a "firewall for output" is implemented as part of a "recording device." In this specification, even when simply describing a "firewall," it is implicitly understood that it is implemented as part of the "recording device and recording software." Also, although Figure 7 illustrates only one "recording device and recording software," It is expected that various monitoring and regulatory conditions will be established in response to the various "information processing devices and information processing software" that correspond to various uses, and that "recording devices and recording software" that correspond to these monitoring and regulatory conditions will be implemented in parallel and / or retrofitted. It is expected that the owners and operators of these numerous "recording devices and recording software" will be responsible for building social consensus, developing systems, implementing systems together, developing verification methods, and handing over the "recording devices and recording software" to the owners and operators as a "complete and complete system."

[0081] Firewalls are installed to protect internal networks from unauthorized access and attacks from the outside, and are often used to monitor inbound traffic and block unauthorized access. However, firewalls can also be used for outbound traffic. For example, they can be used to monitor data sent from an internal network to an external network and block unauthorized activity or connections to malicious computers. They can also be used to prevent confidential information from leaking from internal networks by controlling outbound traffic.

[0082] There are many types of firewalls in use that handle outbound traffic. Some typical examples are: [1] Application Level Gateway (ALG): Application-level gateways understand specific application protocols and, based on that, filter and control outgoing traffic on a per-application basis. For example, they can be used to allow only traffic related to web browsing, or to restrict certain protocols (FTP, SMTP, etc.).

[0083] [2] Proxy Firewall: A proxy firewall is placed between the client and server and relays, filters, and logs traffic. By passing outgoing traffic through a proxy firewall, it is possible to filter and control all outgoing traffic. Proxy firewalls may support protocols such as HTTP, HTTPS, and FTP.

[0084] [3] Next-generation firewall: Next-generation firewalls add functionality to traditional firewalls, such as advanced threat detection, application visibility, and user identification, allowing for more precise filtering, control, and monitoring of outgoing traffic.

[0085] In addition to the image data captured by each camera, a large amount of data containing personal information (privacy) is obtained, including data from IoT sensors, data extracted from captured images through image processing, data obtained through communications with pedestrians' smartphones, and data provided by smart cities and autonomous driving systems. All of this data is likely to be used.

[0086] In the near future, public spaces monitored by various sensors, including cameras, will spread nationwide. The big data extracted from these spaces will be analyzed using artificial intelligence (AI), enabling detailed monitoring of all human and environmental movements and changes. This social system will reliably detect and track street criminals, missing persons, and wanted criminals. It will also predict and detect abnormal behavior, abnormal environments, and terrorism. Dramatic improvements in efficiency are anticipated in many areas of police work (criminal investigation, crime prevention, community safety, patrols, traffic signal control and traffic control, missing person searches, and security to protect citizens from terrorism and disasters). This social system will sense all movements and changes within public spaces, bringing about dramatic improvements in social safety and efficiency.

[0087] In addition to public safety, how can we protect the privacy of ordinary citizens? This is an issue that concerns the very foundation of democracy. In particular, how can we ensure that the misuse of big data generated by the system and the misuse of the results of big data analysis by AI are reliably exposed? This is a particularly challenging issue. In the future, various methods, including the technology and methods provided by this invention, will be proposed and considered, and problems will be identified and solved through social experiments using prototype systems, with the hope that this will be established as a social infrastructure that is acceptable to democratic nations.

[0088] The key to "ensuring that misuse of big data generated by the system and misuse of the results of big data analysis by AI is exposed" is likely to be strict management and recording of "output." To ensure strict management and recording of "output," it is effective to utilize a firewall that supports output (outbound) traffic, or similar technologies and methods.

[0089] In addition to "output," it is possible to manage and record "input," but this raises the following problem. Namely, when big data is constructed from a huge amount of input and is analyzed by an artificial intelligence system, it is often extremely difficult to track the whereabouts of private information. For this reason, this invention focuses on "output." [Example]

[0090] Example 3 shown in Figure 8 is an example of claim 3. Example 3 employs the information output method described in claim 1 or claim 2, characterized in that the recording device and recording software can be operated independently of the information processing device and information processing software.

[0091] In this invention, the administrator, operator, and owner of the "information processing device and information processing software" As a key to uncovering malicious use, we present the concept of "reliable recording of system output," and as a concrete method for achieving this, we provide a system in which "all output" from "information processing devices and information processing software" is monitored and recorded by "recording devices and recording software" operated by a trusted third party.In this context, it is an essential condition that the recording devices and recording software can be operated independently from the information processing devices and information processing software. [Example]

[0092] Example 4 shown in Figure 9 is an example of claim 4. Example 4 employs the information output method described in claims 1 to 3, characterized in that the recording device and recording software record a series of processes using a blockchain.

[0093] Example 4, shown in Figure 9, shows the exchange of information when each camera detects an anomaly. When each camera issues a report or alarm based on the detected anomaly, it is appropriate for each camera (sender) to obtain permission. When the sender receives the trigger of "anomaly detection," "exchange with the outside" begins. Until then, only activity occurs within the sender (inside the AI, etc.), and no information containing privacy information is output to the outside. Stored data includes the date and time of the inquiry, the ID of the provider (AI system), the type, content (abnormal behavior, criminal behavior, etc.) and amount of information provided, and the destination of the alert and report.

[0094] Using a blockchain makes it possible to store data in a verifiable and tamper-proof format. Regularly updating the blockchain (for example, monthly) to a new one is also an effective way to prevent the blockchain from becoming too large and its operating costs from rising. It is also possible that the costs associated with a blockchain record system could be covered by organizations requesting access (such as city halls, police stations, and specialized institutions) paying each time they request access.

[0095] Ethereum, Hyperledger Fabric, etc. are suitable as blockchain platforms for the prototype system.

[0096] The minimum number of servers required to enable a blockchain consensus system depends on the blockchain design and consensus algorithm. Major blockchain consensus algorithms include Proof of Work (PoW), Proof of Stake (PoS), Delegated Proof of Stake (DPoS), and Practical Byzantine Fault Tolerance (PBFT), each of which has different requirements. A sufficient number of nodes and their reliability are important to ensure the reliability and security of the consensus system.

[0097] If the owners, managers, and operators of the "recording devices and recording software" are sufficiently trustworthy, there is no particular need to use blockchain. However, as a practical matter, it is difficult to find or create "sufficiently trustworthy owners, managers, and operators of recording devices and recording software," and to continue to maintain that trust. For this reason, using blockchain is a practical solution. [Example]

[0098] Example 5 shown in Figure 10 is an example of claim 5. Example 5 employs the information output method described in any one of claims 1 to 4, characterized in that the information processing device / information processing software is an artificial intelligence system / artificial intelligence software that receives input of data including privacy information.

[0099] As shown in Figure 10, it is considered appropriate to store the "requested operations" of the "recording device / recording software" and "artificial intelligence system / artificial intelligence software" as a method for storing utilization history for the "data transfer to server, big data analysis using AI" function. One problem with the appropriate use of the vast amount of data generated by analysis by the artificial intelligence (AI) system within the system, and the reliable recording and verification of utilization history, is that the volume of data is so large that the processing details become a black box and are difficult to visualize. In such cases, it is considered effective to record the "output" from the AI ​​system and compare it with pre-defined "good use" and "bad use." The basic process for "retaining utilization history" is as follows.

[0100] (1) The "recording device and recording software" and "artificial intelligence system and software" collect information (image data and analysis results from each camera) from each camera based on predefined rules and build Big Data. (2) The "recording device / recording software" and "artificial intelligence system / artificial intelligence software" analyze big data based on predefined rules and detect anomalies such as "terrorist-specific behavior." (3) If the "Recording Device / Recording Software" or the "Artificial Intelligence System / Artificial Intelligence Software" detects an abnormality, it will send a "Request Code" to the "Recording Device / Recording Software." (4) The "Recording Device / Recording Software" processes the "Request Code" based on pre-defined rules and sends an "Authorization Code" containing the authorized action details to the "Artificial Intelligence System / Artificial Intelligence Software." (5) "Recording devices and software" record the entire process and make it available for public review. (6) The "artificial intelligence system / artificial intelligence software" will take action (such as responding or reporting) based on the content specified in the "authorization code."

[0101] The following are the necessary conditions for a "public space where all movements and changes are monitored" to be suitable for a democratic nation: (a) The information and analysis results obtained will be released from the system and used only for legitimate purposes that have been predetermined through social discussion. (b) This social system is permitted to sense "all movements and changes within the public space," but is obligated never to disclose what it senses or the results of analyzing it outside the system, unless there is a legitimate reason to do so. (c) All utilization history sent outside the system is recorded in a verifiable and tamper-proof form. The utilization history is then available for public verification. [Example]

[0102] Example 6 shown in Figure 11 is an example of claim 6. Example 6 employs the information output method described in any one of claims 1 to 4, characterized in that the information processing device and information processing software are a server system and server software that control cameras.

[0103] In this example, a request for "detecting anomalies and then reporting" is made and set in advance in the "information processing device / information processing software." For example, in "child monitoring," the system automatically monitors the child's entire journey from leaving home, going to school, and returning home. The identification and recognition of the "child" is supplemented (supported) by communication (Bluetooth communication) with the child's smartphone.

[0104] If an abnormality is detected, a report will be sent to the home or security company. The speaker and microphone of the nearest street light unit will also be used to make voice calls and conversations from the home or security company, as well as automatic voice calls and conversations.

[0105] Possible "abnormalities" detected by the street light unit include the following: [1] Detour (to deviate from the designated school route), [2] Passersby call out to you, causing you to stop, [3] Traffic accidents and injuries [Example]

[0106] Example 7 shown in Figure 7 is an example of claims 1, 2, 3, 4, and 5. In Example 7, both input and output traffic of the artificial intelligence system are filtered and controlled by firewalls. "These firewalls" are part of the functions of the "recording device and recording software" defined in this invention. Also, the "artificial intelligence system" is the "information processing device and information processing software" defined in this invention.

[0107] The owner, manager, and operator of the "recording device and recording software" are an independent third-party organization, and the owner, manager, and operator of the "information processing device and information processing software" are also independent third-party organizations. The owner, administrator, and operator of the "recording device and recording software" should preferably be a trusted institution such as a court, and there may be one institution per country.

[0108] Various companies and government agencies may be the owners, managers, and operators of "information processing devices and information processing software." All "information processing devices and information processing software" that handle "big data containing personal information and other privacy-related information" must strictly limit and record its output. Depending on the form of utilization, the target may be "big data that is completely free of personal information and other privacy-related information," or it may be "big data that is completely free of personal information and other privacy-related information." For example, if the present invention is applied to address the threat of artificial intelligence systems related to the "singularity" described below, or the "uselessness of humans as a source of wealth and power due to the intelligence explosion of artificial intelligence," whether or not the input contains privacy-related data is not an essential issue. The primary responsibility for setting conditions, developing an operation method, and proving the effectiveness of a firewall for output traffic to an "information processing device / information processing software" lies with the owner, administrator, and operator of the "information processing device / information processing software." The owner, administrator, and operator of the "recording device / recording software" is responsible for managing and operating the firewall for which conditions have been set, an operation method has been developed, and the effectiveness of which has been proven by the owner, administrator, and operator of the "information processing device / information processing software."

[0109] It is expected that various services and utilization methods will be developed and put into practical use using "information processing devices and information processing software." For each of these, the owner, administrator, and operator of the "information processing devices and information processing software" will be responsible for setting the regulatory conditions, developing the operation method, and proving the effectiveness of the firewall for the output traffic of the "information processing devices and information processing software."

[0110] Artificial intelligence (AI) is seen as "mankind's worst and final invention," and there are concerns about the threat of the singularity (technological singularity). However, there are also many arguments for optimism.

[0111] Prominent figures who have expressed concern about the AI ​​singularity as a threat to humanity include British physicist Stephen Hawking, Microsoft co-founder Bill Gates, and Tesla and SpaceX entrepreneur Elon Musk. These people have issued serious warnings about the potential threat posed by the evolution of AI. They cite the following reasons why AI poses a threat to humanity:

[0112] [a] Loss of control: If AI acquires advanced intelligence, it will become difficult for humans to control it, raising concerns that AI may behave in an unpredictable manner, or may behave in a way that produces results completely different from those humans intended. It has been pointed out that humans may no longer be able to safely control AI. [b] Self-improvement: If AI passes the singularity (technological singularity), there is concern that it may develop a mechanism for improving and multiplying itself. It has been pointed out that once the singularity is passed, evolution could occur at an explosive rate. If this occurs, it has been pointed out that AI could eliminate us humans (Homo sapiens) and become the dominant species.

[0113] On the other hand, some celebrities and experts argue that the AI ​​singularity will not occur and that AI is unlikely to pose a threat to humanity. For example, Microsoft co-founder Paul Allen is skeptical of the arrival of the singularity, claiming that "the singularity isn't near." Other celebrities and experts argue that the AI ​​singularity is likely to occur, but that AI is unlikely to pose a threat to humanity. For example, Ray Kurzweil, who claims that "the singularity is near," argues that the singularity is occurring, but that it will not pose a threat to humanity, and that the fusion of humans and machines will play a complementary role in enhancing human capabilities.

[0114] As mentioned above, both supporters and opponents have expressed various thoughts and outlooks regarding whether the singularity (technological singularity), the point at which AI will attain capabilities on a par with the human brain, will become a reality. However, if you listen carefully to the views of opponents, you will often find that they add the qualification "for the time being," saying, "Based on the current level of AI technology, the singularity will not arrive in the near future."

[0115] Once the singularity is reached, there is concern that AI will self-improve and self-multiply at an explosive rate, quickly growing into a being far superior to human intelligence. At that point, in the process of AI's self-improvement and self-multiplication, its creators, humans, will essentially become unnecessary.

[0116] It is generally believed that "humans" are "things that are physically and autonomously generated from sperm and eggs," and that the "human brain" is also "a physical, material entity that constitutes a neural network." Intelligence is considered to be "the result of information communication within the neural network of the brain as a material entity." The function of neural networks is complex, and at present, we have not yet fully understood it. However, since the formation of "humans" begins with tiny "sperm and eggs" and is "something that is physically and autonomously generated," it is unreasonable to think that this cannot be reproduced artificially.

[0117] Some argue that current AI technology, and by extension its future potential, is far from "realizing the functions of the human brain" and therefore cannot "realize human intelligence." Furthermore, even if "AI that replicates the functions of the human brain" were to be created, there would be little room for "intelligence development" and therefore it may not be possible to achieve "explosive self-improvement." Nevertheless, if we could once replicate the "intelligence" of particularly intelligent individuals (such as Albert Einstein, known as the "greatest physicist of the 20th century," inventor Thomas Edison, also known as the "king of inventions," Galileo Galilei, also known as the "father of modern science," philosopher Aristotle, also known as the "father of all sciences," composer Ludwig van Beethoven, and playwright and poet William Shakespeare), it would be easy to replicate them thousands or even millions of times.

[0118] It is important to note that we do not know when the "breakthrough to the singularity" and "explosively rapid self-improvement" will occur. Furthermore, if the singularity becomes a reality, the impact on humanity could be devastating and irreversible. Considering these factors, "understanding and controlling (regulating)" AI and its potential for development is, to put it mildly, one of the most important challenges facing humanity.

[0119] Throughout human history, people have been the source of wealth and power, and countries and groups with systems and mechanisms that have successfully made people work more efficiently have prevailed. In each era and region, from the Stone Age (hunter-gathering, agriculture), through ancient civilizations, the Middle Ages, the modern era (the Industrial Revolution, the French Revolution, imperialism), and the present day (the Cold War, the Information Revolution, globalization), countries and groups that have succeeded in making people work more efficiently have expanded their influence. For example, in modern times, the reason why democratic and capitalist countries have been more successful in enriching their nations and strengthening their military than totalitarian and communist countries, and the reason why China's socialist market economy has achieved great growth in a globalized environment, is thought to be because they have succeeded in making people (their citizens) work more efficiently.

[0120] When AI surpasses human capabilities in every respect, it is believed that the source of wealth and power will no longer be human. This will be the first time such an event has occurred in the long history of humanity. What will happen at that time is worthy of the utmost interest and attention.

[0121] I believe the root of the threat posed by AI to humanity can be summarized as follows: Throughout human history, countries and systems that use people efficiently have developed. For example, in modern times, democracy and capitalism have gained power and achieved a certain level of development by making effective use of people. However, with the advent of AI after the singularity, the development of science and technology, the invention of new products, and the implementation of industrial and agricultural production will all be possible using AI, potentially eliminating the need for humans. In other words, the efficient use of people will no longer be necessary for countries and systems to develop and be meaningful. The fundamental premise throughout human history that countries and systems that use people more efficiently will eliminate other countries and systems will collapse. What kind of system will dominate the world after this, and will democracies and capitalism be eliminated? I believe a serious threat is looming.

[0122] The possibility that AI will eliminate humanity is an extremely serious threat. However, even before that, the emergence of nations and systems that skillfully use AI rather than humans would also pose a serious threat to democracy. Furthermore, there is concern that this change is irreversible, and that once it occurs, it will be impossible to reverse it.

[0123] With the promise of huge profits for early AI developers, tech companies and researchers from research institutions around the world are pouring huge amounts of research funds into fierce competition in research and development. It appears that the majority of companies and researchers are ignoring the threats to humanity that lie ahead and forging ahead with research and development. Some argue that AI research and development should be temporarily suspended and that effective regulations should be put in place, but this does not seem to be working, with criticism ranging from "they're just trying to buy time because they're behind the curve." This invention is also thought to be useful in detecting signs of runaway AI.

[0124] It is extremely important to prevent AI systems that are still in development from going out of control, which could pose a fatal threat to humanity. To this end, we believe that imposing restrictions based on this invention on the "output" of AI systems will be an essential measure. [1] Carefully considered regulations are strictly applied. [1] Every action is monitored and stored securely in a tamper-proof and verifiable manner. [3] The recorded behavior is analyzed and carefully verified to ensure there are no regulatory gaps. Considering that the workings of AI will become a black box and even developers will be unable to understand its thought processes, careful, continuous, and strict monitoring and regulation of its output will be essential to prevent AI from going out of control.

[0125] In this case, it is particularly significant to make the information output method described in claim 1 or claim 2, characterized in that the recording device and recording software described in claim 3 can be operated independently of the information processing device and information processing software.

[0126] The following are some of the promising information outputs of artificial intelligence systems. First, in various fields such as economics, management, weather, and medicine, big data analysis could predict future trends and patterns. Second, natural language processing, with its ability to understand human language and generate appropriate responses, could be used to output information in applications such as chatbots, machine translation, text summarization, and sentiment analysis. Third, image recognition and computer vision, which are capable of detecting and identifying objects, people, and places from images and video footage, could be used to output extremely powerful information in fields such as autonomous driving, surveillance systems, and medical image diagnosis. Fourth, reinforcement learning, a machine learning technique that learns tasks through interaction with the environment, could be used to output powerful information for decision-making in areas such as robot control, game playing, and financial trading.

[0127] It is conceivable that AI systems could directly access military and financial systems, output information, and directly make decisions and take action. For example, access to military systems could enable autonomous weapons systems to autonomously carry out tasks such as reconnaissance, attack, and defense on the battlefield. Furthermore, military intelligence collection and analysis could enable more effective situational awareness, operational planning, and strategic planning. For example, access to financial systems could enable market data analysis and automated trading to maximize profits. Furthermore, by detecting fraudulent transactions and other illicit activity and strengthening security, it is conceivable that AI systems could protect customer assets and confidential information, thereby aiming for the stability of the entire financial system. However, these assumptions are based on the assumption that AI systems are inherently good and do not anticipate unexpected and / or malicious out-of-control behavior. The information output of artificial intelligence systems is expected to be used in a variety of fields, including medical diagnosis, treatment planning, and treatment implementation, automated driving and traffic efficiency through optimal traffic signal control based on understanding and analysis of traffic conditions, on-demand provision of learning programs and teaching materials based on understanding and analysis of individual students' progress, optimization of operations in the manufacturing and logistics industries based on understanding and analysis of manufacturing processes and logistics chains, and optimization of energy settlement and distribution based on understanding and analysis of the current state of energy utilization.

[0128] Currently, and in the future, artificial intelligence systems will be required to output a variety of information for a variety of purposes in a variety of fields. It is believed that the entity attempting to output such information should bear the primary responsibility for proving, developing, and bearing the costs of building a social consensus on the appropriateness of each purpose, and for determining whether the information output is in line with that purpose. In particular, there is a need to develop, prove, and provide verification methods to check whether the information output contains fraudulent content. Examples of fraudulent content include malicious viruses (spyware, ransomware, Trojan horses, worms, etc.) that can be embedded in program code (such as computer applications).

[0129] [Industrial Applicability]

[0130] This invention provides a method for protecting the privacy of ordinary citizens by reliably monitoring and controlling the output of an artificial intelligence system that receives and analyzes big data composed of privacy-containing data such as security camera images, credit card usage history, smartphone and smartwatch usage history and location information, and various sensor information (biomotor information, environmental monitoring information).It also provides a method for detecting signs of runaway AI systems due to the singularity, etc.

Claims

1. An information output method characterized in that an information processing device / information processing software obtains permission for information output from a recording device / recording software before outputting information.

2. 2. The information output method according to claim 1, wherein the recording device and recording software have a firewall function for monitoring, controlling and recording the output.

3. 3. The information output method according to claim 1, wherein the recording device and recording software can be operated independently of the information processing device and information processing software.

4. An information output method according to claims 1 to 3, characterized in that the recording device and recording software record a series of processes using a blockchain.

5. An information output method according to any one of claims 1 to 4, characterized in that the information processing device / information processing software is an artificial intelligence system / artificial intelligence software that receives input of data including privacy information.

6. The information output method according to any one of claims 1 to 4, characterized in that the information processing device and information processing software are a server system and server software that control the camera.

Images