User privacy protection methods based on large language model applications

Abstract

This application discloses a user privacy protection method for a large language model-based application. The method includes: for a large language model-based application on a terminal, when the remaining battery power is greater than a specific threshold, setting permissions for the large language model's inference, training, parameter sharing, and dataset sharing, and waiting for user confirmation. Specifically, the inference permission setting includes prompting the user to enable inference on the local machine, on the network side, or both simultaneously; the training permission setting includes prompting the user to enable training on the local machine, on the network side, or both simultaneously; the parameter sharing permission setting includes prompting the user to confirm whether model parameters can be shared with the network side; and the dataset sharing permission setting includes prompting the user to confirm whether raw data can be shared with the network side, raw data cannot be shared with the network side, inference data can be shared with the network side, or inference data cannot be shared with the network side. When the remaining battery power is less than or equal to the specific threshold, the large language model is not applied, and its permission settings are not enabled.

Description

Technical Field

[0001] This invention relates to user privacy protection technology, and more specifically, to a method and apparatus for user privacy protection based on large language model applications. Background Technology

[0002] With advancements in chip computing power and artificial intelligence algorithms, large language models (LLMs) are now used in mobile devices for text, speech, and image processing. Here, a large language model generally refers to a deep learning model trained on massive amounts of text data, capable of generating natural language text or understanding the meaning of spoken text. Large language models can also be simply called large models, generally referring to AI models with a large number of parameters. Mobile-based large language models can have parameter sets in the hundreds of millions or even billions, capable of handling some simple AI applications. However, for more complex tasks, such as image processing or audio / video processing, limitations imposed by the model and mobile device hardware necessitate processing by the large language model on the network side. The application of large language models, due to the complexity of the processing, can impact the terminal's power consumption and user experience. Furthermore, network-based large language model applications require users to share data for inference or training, raising concerns about user data security. Therefore, a comprehensive consideration of user experience, user data security, and the benefits of using large language models is necessary.

[0003] Currently, research on introducing large language model technology into terminal data processing is in its early stages, and there are many optimization issues that need to be addressed. Summary of the Invention

[0004] In view of the above, the present invention provides the following technical solution:

[0005] A method for protecting user privacy, applied to a large language model application on a terminal, is characterized by comprising: for a large language model-based application on the terminal, when the remaining battery power is greater than a specific threshold, setting permissions for the large language model's inference, training, parameter sharing, and dataset sharing, and waiting for user confirmation. Specifically, the inference permission setting includes prompting the user to enable inference on the local machine, on the network side, or both simultaneously; the training permission setting includes prompting the user to enable training on the local machine, on the network side, or both simultaneously; the parameter sharing permission setting includes prompting the user to confirm whether model parameters can be shared with the network side; and the dataset sharing permission setting includes prompting the user to confirm whether raw data can be shared with the network side, raw data cannot be shared with the network side, inference data can be shared with the network side, and inference data cannot be shared with the network side. When the remaining battery power is less than or equal to the specific threshold, the large language model is not applied, and its permission settings are not enabled.

[0006] This method includes the application of large language models, including at least the training, learning, and inference of large language models based on text summarization, language translation, audio optimization, or video optimization.

[0007] The method further includes, on the network side, at least: base stations, multi-access edge servers, core networks, or artificial intelligence servers.

[0008] The method further includes raw data, including at least text, image, and video data used for large language model inference and training.

[0009] A user privacy protection processing device applied to large language model processing on a terminal is characterized by comprising: for a large language model-based application on the terminal, when the remaining battery power is greater than a specific threshold, setting permissions for large language model inference, large language model training, large language model parameter sharing, and dataset sharing, and waiting for user confirmation. Specifically, the large language model inference permission setting includes prompting the user to enable large language model inference on the local machine, on the network side, or simultaneously on both the local machine and the network side; the large language model training permission setting includes prompting the user to enable large language model training on the local machine, on the network side, or simultaneously on both the local machine and the network side; the large language model parameter sharing permission setting includes prompting the user to indicate whether model parameters can be shared with the network side; and the dataset sharing permission setting includes prompting the user to indicate whether raw data can be shared with the network side, raw data cannot be shared with the network side, inferred data can be shared with the network side, and inferred data cannot be shared with the network side. When the remaining battery power is less than or equal to the specific threshold, the large language model is not applied, and its permission settings are not enabled.

[0010] As can be seen from the above technical solution, compared with the prior art, the embodiments of the present invention disclose a method for protecting user privacy, applied to a large language model application on a terminal. The method includes: for a large language model-based application on the terminal, when the remaining battery power is greater than a specific threshold, the terminal sets permissions for inference, training, parameter sharing, and dataset sharing of the large language model, and waits for user confirmation. Specifically, the inference permission setting includes prompting the user to enable inference on the local machine, on the network side, or simultaneously on both the local machine and the network side; the training permission setting includes prompting the user to enable training on the local machine, on the network side, or simultaneously on both the local machine and the network side; the parameter sharing permission setting includes prompting the user to indicate whether model parameters can be shared with the network side; and the dataset sharing permission setting includes prompting the user to indicate whether raw data can be shared with the network side, raw data cannot be shared with the network side, inferred data can be shared with the network side, and inferred data cannot be shared with the network side. When the remaining battery power is less than or equal to the specific threshold, the terminal does not apply the large language model and does not enable its permission settings. Attached Figure Description

[0011] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0012] Figure 1 This is a flowchart of a user privacy protection method for a terminal large language model application disclosed in an embodiment of the present invention;

[0013] Figure 2 This is a schematic diagram of a user privacy protection method for a terminal large language model application disclosed in an embodiment of the present invention;

[0014] Figure 3 This is a functional diagram illustrating the user privacy protection method for a large language model application in a terminal disclosed in an embodiment of the present invention. Detailed Implementation

[0015] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0016] Figure 1 This is a flowchart illustrating a user privacy protection method for a terminal large language model application disclosed in an embodiment of the present invention. See also... Figure 1 As shown, the flowchart includes the following steps:

[0017] Step 101: The terminal receives a request from the user regarding the application of the large language model, such as image processing, language translation, or optimization of wireless signal reception;

[0018] Step 102: The terminal determines whether its remaining battery power is greater than a specific threshold, such as 20%. Considering that the application of the large language model is relatively power-intensive, the terminal needs to ensure that the application of the large language model will not affect normal use. If the remaining battery power is less than or equal to the specific threshold in actual use, the terminal will not apply the large language model or enable its permission settings (Step 103). This also means that the user will not use the large language model for related problem processing. If the remaining battery power is greater than the specific threshold in actual use; proceed to Step 104: enable permission settings for the large language model inference, large language model training, large language model parameter sharing, and dataset sharing respectively, and wait for user confirmation. The inference permission settings for the large language model include prompting the user to enable inference on the local machine, the network side, or both simultaneously; the training permission settings include prompting the user to enable training on the local machine, the network side, or both simultaneously; the parameter sharing permission settings include prompting the user to indicate whether model parameters can be shared with the network side; and the dataset sharing permission settings include prompting the user to indicate whether raw data can be shared with the network side, raw data cannot be shared with the network side, inference data can be shared with the network side, and inference data cannot be shared with the network side. Users select the corresponding permissions based on these settings, and the terminal performs the corresponding large language model application based on the user's permission settings. For example, if the user selects that raw data cannot be shared with the network side, the terminal cannot transmit the raw data to the network side. If the user selects that model parameters (model structure, number of layers, gradient information, etc.) can be shared with the network side, the terminal can transmit this data to the network side for related processing applications, such as model optimization.

[0019] Additionally, users can pre-configure the relevant large language model application permissions for each application on the terminal. The terminal then performs the relevant large language model processing applications based on its battery level and the user-defined permissions.

[0020] Figure 2This is a schematic diagram of a user privacy protection method for a terminal large language model application disclosed in an embodiment of the present invention. The embodiment includes a terminal and a network. The embodiment includes the following steps:

[0021] Step 0: The terminal receives a request to apply the large language model;

[0022] Step 1: The terminal determines whether its remaining battery power is greater than a specific threshold (e.g., 20%). If the above conditions are met, proceed to Step 2.

[0023] Step 2: The terminal enables permission settings for inference, training, parameter sharing, and dataset sharing of the large language model, and waits for user confirmation.

[0024] Step 3: The terminal transmits relevant data;

[0025] Step 4: Perform inference on the large language model on the network side;

[0026] Step 5: The network side returns the inference results.

[0027] This embodiment balances terminal power consumption and user privacy protection, and can effectively optimize various subsequent applications on the terminal (such as text, image, audio and video, and wireless signal processing), thereby improving the user experience.

[0028] For the foregoing method embodiments, in order to simplify the description, they are all described as a series of actions. However, those skilled in the art should understand that the present invention is not limited to the described order of actions, because according to the present invention, some steps can be performed in other orders or simultaneously. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions and modules involved are not necessarily essential to the present invention.

[0029] The methods described in the above-disclosed embodiments of the present invention are detailed in terms of their specific implementation. The methods of the present invention can be implemented using various forms of devices. Therefore, the present invention also discloses a device, and specific embodiments are given below for detailed description.

[0030] Figure 3This invention discloses a privacy protection processing unit for a large language model application. The data receiving module 301 receives various types of data, such as raw data to be processed by the large language model, network signaling, and internal terminal instructions. The data processing module 302 is primarily used for privacy protection decisions by the large language model. Module 303 is a data sending module, which sends internal instructions to the terminal, interaction signaling with the network side, and raw data. The signaling processing module 304 is a network signaling parsing and encapsulation module.

[0031] This embodiment describes a user privacy protection processing device applied to large language model processing on a terminal. Its features include: for a large language model-based application on the terminal, when the remaining battery power is greater than a specific threshold, setting permissions for large language model inference, large language model training, large language model parameter sharing, and dataset sharing, and waiting for user confirmation. Specifically, the large language model inference permission setting includes prompting the user to enable large language model inference on the local machine, on the network side, or simultaneously on both the local machine and the network side; the large language model training permission setting includes prompting the user to enable large language model training on the local machine, on the network side, or simultaneously on both the local machine and the network side; the large language model parameter sharing permission setting includes prompting the user to indicate whether model parameters can be shared with the network side; and the dataset sharing permission setting includes prompting the user to indicate whether raw data can be shared with the network side, raw data cannot be shared with the network side, inferred data can be shared with the network side, and inferred data cannot be shared with the network side. When the remaining battery power is less than or equal to the specific threshold, the terminal does not apply the large language model and does not enable its permission settings.

[0032] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to the method section.

[0033] It should also be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0034] The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein can be implemented directly by hardware, a software module executed by a processor, or a combination of both. The software module can be located in random access memory (RAM), main memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, removable disk, CD-ROM, or any other form of storage medium known in the art.

[0035] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for protecting user privacy, applied to a large language model application on a terminal, characterized in that, include: For a large language model-based application in a terminal, the terminal receives a user's request for the large language model application, determines whether the terminal's remaining battery power is greater than a specific threshold, and if the remaining battery power is greater than the specific threshold, the terminal sets permissions for the large language model's inference, training, parameter sharing, and dataset sharing, and waits for user confirmation. Specifically, the inference permission settings include prompting the user to enable inference on the local machine, on the network side, or both simultaneously; the training permission settings include prompting the user to enable training on the local machine, on the network side, or both simultaneously; the parameter sharing permission settings include prompting the user to indicate whether model parameters can be shared with the network side; and the dataset sharing permission settings include prompting the user to indicate whether raw data can be shared with the network side, raw data cannot be shared with the network side, inferred data can be shared with the network side, and inferred data cannot be shared with the network side. If the remaining battery power is less than or equal to the specific threshold, the terminal neither applies the large language model nor enables its permission settings. Users select the corresponding permissions based on the above settings, and the terminal applies the corresponding large language model based on the user's permission settings.

2. The method for protecting user privacy according to claim 1, characterized in that, The aforementioned applications based on large language models include at least: Training, learning, and inference of large language models used for text summarization, language translation, audio optimization, or video optimization.

3. The method for protecting user privacy according to claim 1, characterized in that, The network side, as described, includes at least: Base stations, multi-access edge servers, core networks, and artificial intelligence servers.

4. The method for protecting user privacy according to claim 1, characterized in that, The raw data mentioned above includes at least: Text, image, and video data used for reasoning and training of large language models.

5. A user privacy protection processing device, applied to large language model processing on a terminal, characterized in that, include: For a large language model-based application in a terminal, when the remaining battery power is greater than a certain threshold, the terminal sets permissions for the large language model's inference, training, parameter sharing, and dataset sharing, and waits for user confirmation. Specifically, the inference permission settings include prompting the user to enable inference on the local machine, the network side, or both simultaneously; the training permission settings include prompting the user to enable training on the local machine, the network side, or both simultaneously; the parameter sharing permission settings include prompting the user to confirm whether model parameters can be shared with the network side; and the dataset sharing permission settings include prompting the user to confirm whether raw data can be shared with the network side, raw data cannot be shared with the network side, inference data can be shared with the network side, and inference data cannot be shared with the network side. When the remaining battery power is less than or equal to the certain threshold, the terminal neither applies the large language model nor enables its permission settings.

Images