Method and system for information protection using alternative text generation

Abstract

The present invention relates to an information protection method and system using alternative text generation, which allows a user to obtain a desired result while preventing direct exposure of information to a company providing an AI model by generating alternative text that preserves the information the user wishes to protect while maintaining it as is, and inputting the prompt. The present invention comprises: a user terminal (100) that receives a prompt, allows the user to select important information to be protected and not exposed within the prompt, and when important information is selected, transmits the selected important information to a personal information protection server (200) and receives the result; and a personal information protection server (200) that, when a prompt is input from the user terminal (100), analyzes the input prompt to generate alternative text for the important information to be protected and transmits a prompt in which user information is not exposed to a commercial LLM API server (300), and when response data for the prompt transmitted from the commercial LLM API server (300) is received, restores the data to be transmitted to the user terminal (100) in conjunction with the prompt input from the user terminal (100) and transmits it. The present invention provides an information protection system using alternative text generation, characterized by comprising: a commercial LLM API server (300) that generates response data for a prompt transmitted from the above-mentioned personal information protection server (200) and responds.

Description

Technology Field

[0001] The present invention relates to an information protection method and system using alternative text generation, and more specifically, to an information protection method and system using alternative text generation that can protect one's personal information or information one does not wish to be exposed, which is included in the prompts entered by a user when using a commercial LLM model. Background Technology

[0003] There is growing anxiety that companies are storing and using user-inputted data without authorization during the process of using AI models operated by companies.

[0004] In addition, there is also anxiety regarding the AI ​​learning the sentence structures and writing styles used by creators, as these are products belonging to the creator's unique creative process.

[0005] Furthermore, when companies use AI models, the exposure of their technical details to the AI ​​creates anxiety that other companies might generate data based on their ideas if they use the AI ​​in the future.

[0006] Nevertheless, due to the problem that results are not generated properly if necessary information is not entered, users are forced to enter all their information and use the system.

[0007] Meanwhile, with the passage of the three data laws on January 9, 2020, measures to de-identify personal information for the utilization of big data are emerging as a mandatory requirement, and de-identified personal information must be provided.

[0008] According to recent trends toward a data economy, the utilization of big data is inevitable, and the resulting issues regarding personal information infringement are on the rise. To address this, the provision of personal information anonymization solutions is essential; however, as the manual processing methods offered by existing solutions have limitations, there is a need for an intelligent processing system that improves upon them.

[0009] The security market, including personal information anonymization solutions, is currently in its early stages. However, as a solution applicable to all fields utilizing data, its value in the market is expected to continue to rise.

[0010] Personal information de-identification solutions are an essential requirement for the development of the big data industry, and personal information de-identification refers to de-identification processes such as pseudonymization, aggregation, data deletion, data categorization, and data masking when utilizing personal information.

[0011] The personal information de-identification solution aims to safely and rapidly de-identify such personal information through automated process procedures equipped with artificial intelligence, in accordance with the personal information de-identification guidelines. Prior art literature

[0013] Prior Art 1: Korean Registered Patent No. 10-2627819 (Published Jan. 23, 2024) Prior Art 2: Korean Published Patent No. 10-2021-0110765 (Published Sep. 9, 2021) The problem to be solved

[0014] The present invention is proposed to solve the aforementioned conventional problems, and aims to provide an information protection method and system using alternative text generation that allows a user to obtain a desired result while preventing the direct exposure of information to a company providing an AI model by generating alternative text that preserves the information the user wishes to protect while maintaining it as is, and inputting this text as a prompt. means of solving the problem

[0016] The present invention, for achieving the above-mentioned purpose, provides an information protection system using alternative text generation, characterized by comprising: a user terminal (100) that receives a prompt, allows the user to select important information to be protected and not exposed within the prompt, and when important information is selected, transmits the selected important information to a personal information protection server (200) and receives the result; a personal information protection server (200) that, when a prompt is input from the user terminal (100), analyzes the input prompt to generate alternative text for the important information to be protected and transmits a prompt in which user information is not exposed to a commercial LLM API server (300), and when response data for the prompt transmitted from the commercial LLM API server (300) is received, restores the data to be transmitted to the user terminal (100) in conjunction with the prompt input from the user terminal (100) and transmits it; and a commercial LLM API server (300) that generates response data for the prompt transmitted from the personal information protection server (200) and responds.

[0017] Here, the user terminal (100) is characterized by being composed of a PC, a smartphone, or a smart pad.

[0018] The personal information protection server (200) is characterized by comprising: a prompt storage unit (220) in which a prompt transmitted from a user terminal (100) is stored; a prompt analysis unit (230) that analyzes the prompt transmitted from the user terminal (100) to analyze the section (area) in which protection information is selected; an alternative text generation unit (240) that generates alternative text for the content to be protected among the prompts analyzed by the prompt analysis unit (230); and an alternative prompt conversion unit (250) that converts the alternative text generated by the alternative text generation unit (240) into an alternative prompt.

[0020] In addition, to achieve the above-mentioned purpose, the present invention comprises the following steps: inputting an original prompt through a prompt receiving unit (120) at a user terminal (100) (S100); selecting important information requiring protection by a protection information selection unit (140) within the original prompt (S110); transmitting the prompt with the selected important information from the user terminal (100) to a personal information protection server (200) (S120); the personal information protection server (200) analyzing the prompt to generate alternative text for the prompt with the selected important information (S130); the personal information protection server (200) converting the generated alternative text into an alternative prompt (S140); and transmitting the alternative prompt to a commercial LLM API server (300) (S150). The above personal information protection server (200) includes the step of restoring the response data to the original prompt in conjunction with the response data when response data is received from the above commercial LLM API server (300) (S160) (S170); and the above personal information protection server (200) transmits the restored response data to the user terminal (100) (S180).

[0021] Here, the step (S110) in which the important information is selected is characterized by being one or more of the following: a user selection method in which a specific section of the prompt is selected by a mouse or a specific string is directly entered on the user terminal to designate the important information, and an automatic selection method in which the important personal information within the prompt entered on the user terminal (100) is automatically detected and protected so as not to be exposed, such as the user's personal information including name, address, phone number, and account. Effects of the invention

[0023] The present invention, constructed in this manner, has the following effects.

[0024] First, by configuring the system so that personal information such as names, addresses, accounts, and contact details entered by users in the prompt, as well as sensitive information, internal company information, and confidential information that users do not wish to expose, are not directly transmitted to commercial LLM APIs but are instead sent externally only after being replaced with alternative text generated by a privacy protection server, the risk of personal information leakage and information accumulation due to model training can be fundamentally prevented.

[0025] Second, by enabling functions such as generating alternative text or determining the protection level of important information to be performed on an independent AI model or server separate from commercial LLM, the information protection logic can operate within the company's internal network or in a separate security environment without being dependent on external companies, thereby allowing the company to directly control and manage the personal information protection process. Brief explanation of the drawing

[0027] FIG. 1 is a block diagram illustrating an embodiment of an information protection system using alternative text generation according to the present invention. FIG. 2 is a block diagram illustrating an embodiment of a user terminal of an information protection system using alternative text generation shown in FIG. 1. FIG. 3 is a block diagram illustrating an embodiment of a personal information protection server of an information protection system using alternative text generation shown in FIG. 1. FIG. 4 is a flowchart illustrating an example of an information protection method using alternative text generation according to the present invention. Specific details for implementing the invention

[0028] Hereinafter, a preferred embodiment according to the present invention will be described in detail with reference to the attached drawings.

[0029] Embodiments of the present invention may be modified in various forms, and the scope of the present invention should not be interpreted as being limited to the embodiments described in detail below. These embodiments are provided to more fully explain the present invention to those with average knowledge in the art. Accordingly, the shapes of elements in the drawings may be exaggerated to emphasize clearer explanations. It should be noted that in each drawing, identical components may be depicted with the same reference numeral. Detailed descriptions of known functions and configurations that are deemed to unnecessarily obscure the essence of the present invention are omitted.

[0031] FIG. 1 is a block diagram illustrating an embodiment of an information protection system using alternative text generation according to the present invention, FIG. 2 is a block diagram illustrating an embodiment of a user terminal of an information protection system using alternative text generation shown in FIG. 1, and FIG. 3 is a block diagram illustrating an embodiment of a personal information protection server of an information protection system using alternative text generation shown in FIG. 1.

[0032] An embodiment of an information protection system using alternative text generation according to the present invention is composed of a user terminal (100), a personal information protection server (200), and a commercial LLM API server (300), as shown in FIG. 1.

[0033] The user terminal (100) receives the result after the user inputs a prompt, selects important information to be protected that is not to be exposed within the prompt, and transmits the input prompt to the personal information protection server (200).

[0034] This user terminal (100) can be configured as a PC or a smart device (smartphone, smart pad, etc.), and the user can select information they do not want to expose from background information entered on the display of the terminal. In this case, it is not necessary to specify an exact range, but the protected information is selected by entering it universally or by setting a range.

[0035] The personal information protection server (200) generates a conversion prompt, generates alternative text, and performs integration with the commercial LLM API server (300), and acts as a proxy to prevent important information from being directly transmitted to the external commercial LLM API server (300).

[0036] This personal information protection server (200) is equipped with a proprietary AI model, which is a distinct AI model unlike Google or Open AI, and converts and generates user information into different text when a prompt is delivered.

[0037] In addition, if a user-specified part is essential for achieving good results, it informs the user of this again or suggests a method to protect it in another way.

[0038] In addition, if a result is delivered from the AI ​​model's API, the text information after conversion is found in the result, converted back to its original state, and generated.

[0039] The commercial LLM API server (300) is a commercial API (Application Programming Interface) provided by Google or Open AI.

[0041] Meanwhile, the user terminal (100) may be configured as an embodiment as shown in FIG. 2 and is composed of a communication unit (110), a prompt receiving unit (120), a display (130), a protection information selection unit (140), a protection prompt transmission request unit (150), a power supply unit (160), and a control unit (170).

[0042] Here, the communication unit (110) communicates with the personal information protection server (200).

[0043] The prompt receiver (120) receives a prompt input from a user's input means (mouse, keyboard, etc.).

[0044] The display (130) displays prompts input by the user and results responded to by the personal information protection server (200). These prompts and results include text, files, and video information.

[0045] The protection information selection unit (140) selects (designates) information that needs protection from the user. This information that needs protection is intended to select information that may expose text data, personal information or surrounding information in images.

[0046] This selection of protection information can be classified into user-defined selection and automatic detection selection, where user-defined selection allows the user to specify important information by selecting a specific section of the prompt with a mouse or directly entering a specific string.

[0047] The automatic detection selection is intended to automatically detect and protect key personal information within a prompt entered from a user terminal (100). After the user selects the information to be protected, the user's personal information, such as name, address, phone number, and account, may be automatically protected so as not to be exposed, or a method may be used to check whether the personal information is exposed through a display (130).

[0048] The protection prompt transmission request unit (150) is requested to transmit the user-entered prompt to the personal information protection server (200) after the protection information selection unit (140) completes the selection of protection information.

[0049] The power supply unit (130) supplies the necessary power to the user terminal (100).

[0050] The control unit (170) controls the communication unit (110), the prompt receiving unit (120), the display (130), the protection information selection unit (140), the protection prompt transmission request unit (150), and the power supply unit (160), and controls the display (120) to display whether personal information is exposed in the automatic detection selection function so that the user can check.

[0052] And the personal information protection server (200) can be configured as an embodiment as shown in FIG. 3, and is composed of a communication unit (210), a prompt storage unit (220), a prompt analysis unit (230), an alternative text generation unit (240), an alternative prompt conversion unit (250), an LLM response receiving unit (260), a response data restoration unit (270), a power supply unit (280), and a control unit (290).

[0053] Here, the communication unit (210) communicates with the user terminal (100) and the commercial LLM API server (300).

[0054] The prompt storage unit (220) stores the prompt transmitted from the user terminal (100).

[0055] The prompt analysis unit (230) analyzes the prompt transmitted from the user terminal (100) and analyzes the section (area) where protection information is selected.

[0056] The alternative text generation unit (240) generates alternative text for the content to be protected (such as actual text or text in an image) among the prompts analyzed by the prompt analysis unit (230). This alternative text is stored in the prompt storage unit (220) in conjunction with the identification information of the user terminal (100) and the prompt transmitted from the user terminal (100) so that it can be used upon recovery.

[0057] The alternative prompt conversion unit (250) converts the alternative text generated by the alternative text generation unit (240) into an alternative prompt. The converted alternative prompt is then transmitted to the commercial LLM API server (300).

[0058] The LLM response receiving unit (260) receives the result of the response from the commercial LLM API server (300).

[0059] The response data restoration unit (270) restores the response data to be transmitted to the user terminal (100) by linking the response result received from the LLM response receiving unit (260) with the prompt transmitted from the user terminal (100) and the converted alternative prompt stored in the prompt storage unit (220).

[0060] Meanwhile, the prompt storage unit (220), prompt analysis unit (230), alternative text generation unit (240), alternative prompt conversion unit (250), LLM response receiving unit (260), and response data restoration unit (270) can be operated by configuring a separate AI module for the personal information protection server (200).

[0061] The power supply unit (280) supplies the necessary power to the personal information protection server (200).

[0062] The control unit (290) controls the prompt storage unit (220), the prompt analysis unit (230), the alternative text generation unit (240), the alternative prompt conversion unit (250), the LLM response receiving unit (260), the response data restoration unit (270), and the power supply unit (280).

[0064] FIG. 4 is a flowchart illustrating an example of an information protection method using alternative text generation according to the present invention.

[0065] An example of an information protection method using alternative text generation according to the present invention is shown in FIG. 4, where a prompt is input to a user terminal (100) (S100).

[0066] Then, important information requiring protection within the prompt is selected by user selection or automatic detection (S110). This includes a user selection method in which the user selects a specific section of the prompt with a mouse or directly inputs a specific string to specify important information, and an automatic selection method in which important personal information within the prompt entered by the user terminal (100) is automatically detected and protected. After the user selects the information to be protected, the user's personal information, such as name, address, phone number, and account, is automatically protected so as not to be exposed, or the user can check whether the personal information is exposed through the display (130).

[0067] Next, a prompt in which protection information is selected is transmitted from the user terminal (100) to the personal information protection server (200) (S120).

[0068] The personal information protection server (200) stores and analyzes the transmitted prompt, and generates alternative text so that the important information for which protection is requested can be protected (S130).

[0069] And the personal information protection server (200) converts the transmitted prompt into an alternative prompt in which important information is protected (S140).

[0070] Next, the personal information protection server (200) sends an alternative prompt to the commercial LLM API server (300) (S150).

[0071] Accordingly, the commercial LLM API server (300) generates response data for the prompt transmitted from the personal information protection server (200) and transmits it to the personal information protection server (200), and accordingly, the personal information protection server (200) receives response data from the commercial LLM API server (300) (S160).

[0072] The personal information protection server (200) restores the response data by linking it with the original information for the responded data (S170).

[0073] And the personal information protection server (200) transmits the restored data to the user terminal (100).

[0075] The embodiment of the present invention described above is merely illustrative, and those skilled in the art will readily understand that various modifications and equivalent alternative embodiments are possible therefrom. Therefore, it will be understood that the present invention is not limited only to the forms mentioned in the above detailed description. Accordingly, the true technical scope of protection of the present invention should be determined by the technical spirit of the appended claims. Furthermore, the present invention should be understood to include all modifications, equivalents, and substitutions within the spirit and scope of the present invention as defined by the appended claims. Explanation of the symbols

[0077] 100 : User terminal 110 : Communication unit 120 : Prompt receiver 130 : Display 140: Protection Information Selection Section 150: Protection Prompt Transmission Request Section 160: Power supply unit 170: Control unit 200 : Personal Information Protection Server 210 : Communications Department 220 : Prompt storage unit 230 : Prompt analysis unit 240 : Alternative text generation unit 250 : Alternative prompt conversion unit 260: LLM Response Receiver 270: Response Data Recovery Unit 280: Power supply unit 290: Control unit 300 : Commercial LLM API Server

Claims

Claim 1 An information protection system using alternative text generation, characterized by comprising: a user terminal (100) that receives a prompt, allows the user to select important information to be protected and not exposed within the prompt, and when important information is selected, transmits the selected important information to a personal information protection server (200) and receives the result; a personal information protection server (200) that, when a prompt is entered from the user terminal (100), analyzes the entered prompt to generate alternative text for the important information to be protected and transmits a prompt in which user information is not exposed to a commercial LLM API server (300), and when response data for the prompt transmitted from the commercial LLM API server (300) is received, restores the data to be transmitted to the user terminal (100) in conjunction with the prompt entered from the user terminal (100) and transmits it; and a commercial LLM API server (300) that generates response data for the prompt transmitted from the personal information protection server (200) and responds. Claim 2 An information protection system using alternative text generation, characterized in that, in claim 1, the user terminal (100) is configured as a PC, a smartphone, or a smart pad. Claim 3 An information protection system using alternative text generation according to claim 1, wherein the personal information protection server (200) comprises a prompt storage unit (220) in which a prompt transmitted from the user terminal (100) is stored, a prompt analysis unit (230) that analyzes the prompt transmitted from the user terminal (100) to analyze the section (area) in which protection information is selected, an alternative text generation unit (240) that generates alternative text for the content to be protected among the prompts analyzed by the prompt analysis unit (230), and an alternative prompt conversion unit (250) that converts the alternative text generated by the alternative text generation unit (240) into an alternative prompt. Claim 4 A step (S100) in which an original prompt is input through a prompt receiving unit (120) at a user terminal (100); a step (S110) in which important information requiring protection is selected by a protection information selection unit (140) within the original prompt; a step (S120) in which a prompt with selected important information is transmitted from the user terminal (100) to a personal information protection server (200); a step (S130) in which the personal information protection server (200) analyzes the prompt and generates alternative text for the prompt with selected important information; a step (S140) in which the personal information protection server (200) converts the generated alternative text into an alternative prompt; a step (S150) in which the personal information protection server (200) transmits the alternative prompt to a commercial LLM API server (300); a step (S170) in which, when response data is received from the commercial LLM API server (300) at the personal information protection server (200), the response data is restored to the response data by linking it with the original prompt. A method for protecting information using alternative text generation, characterized by including the step (S180) of the above personal information protection server (200) transmitting the restored response data to the above user terminal (100). Claim 5 In claim 4, the step (S110) of selecting important information is characterized by being one or more of the following: a user selection method in which important information is designated by selecting a specific section of the prompt with a mouse or by directly inputting a specific string of characters at the user terminal; and an automatic selection method in which important personal information within the prompt entered at the user terminal (100) is automatically detected and protected so as not to be exposed, thereby automatically protecting information such as the user's personal information, such as name, address, phone number, and account.

Images