A VPN fast auto-negotiation connection method and device

By building a local knowledge base and using multi-threaded retrieval and LLM large models, VPN parameters are automatically adjusted, solving the problems of complex VPN connection configuration and difficult debugging, achieving fast self-negotiation connection, and improving the convenience and accuracy of connection.

CN120675907BActive Publication Date: 2026-11-10SHANDONG YOU INTERNET OF THINGS CO LTD
View PDF 2 Cites 0 Cited by

Patent Information

Application Number
CN202510952484.8
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2025-07-10
Publication Date
2026-11-10
Estimated Expiration
2045-07-10

AI Technical Summary

Technical Problem

Existing VPN connections are complex to configure, have numerous parameters, and are difficult to debug. They cannot automatically adjust parameters based on error logs, negotiate connections automatically, or proactively accumulate knowledge bases, which affects the model's generalization ability and connection accuracy.

Method used

By building a local knowledge base, including an encoded dictionary and a RAG vector knowledge base, and combining multi-threaded retrieval with a trained LLM large model, VPN parameters are automatically adjusted to achieve self-negotiation connection.

Benefits of technology

It reduces the complexity of VPN configuration, improves the convenience and reliability of connections, and continuously improves connection accuracy through self-learning.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN120675907B_ABST
    Figure CN120675907B_ABST
Patent Text Reader

Abstract

This invention belongs to the field of VPN connection technology and provides a method and apparatus for fast self-negotiation VPN connection. The technical solution involves determining whether the VPN to be detected is experiencing connection problems. If so, a constructed local knowledge base is invoked, including an encoded dictionary and a RAG vector knowledge base. A first solution is obtained through multi-threaded retrieval using the encoded dictionary, and a second solution is obtained by combining the RAG vector knowledge base and a trained LLM model. The first and second solutions are evaluated to obtain a final solution. It is then determined whether the VPN, adjusted according to the final solution, is still experiencing connection problems. If so, the previously output second and final solutions are used as prompts and input again into the trained LLM model until a solution that restores VPN connection is obtained is found. Finally, the VPN anomaly log and the corresponding solution are updated to the local knowledge base. This significantly reduces the configuration difficulty of VPN connections and improves the convenience of connection.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention belongs to the field of VPN connection technology, and particularly relates to a VPN fast auto-negotiation connection method and apparatus. Background Technology

[0002] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art.

[0003] With the continuous development of network technology, Virtual Private Networks (VPNs) have been widely used in various fields, providing users with secure and reliable network connection services. However, traditional VPN connection methods have some problems, such as complex configuration, numerous parameters, and difficulty in debugging, which bring inconvenience to users.

[0004] Currently, various error logs frequently appear during VPN connections. Users need to manually debug and modify configuration parameters based on these error logs, which is not only time-consuming and labor-intensive but also requires users to have professional network knowledge. Furthermore, the types of error logs are numerous, and different error logs require different solutions, posing a significant challenge to users.

[0005] Currently, some methods for addressing the complexity of VPN connection configuration have been disclosed. For example, some methods utilize the textual semantic features of router log data from deep learning neural network models to remotely assess router fault types through log data analysis. However, these methods still rely on remotely transmitting logs to a large server model and using the network for diagnosis and evaluation. They lack data augmentation methods specific to VPN connection scenarios, failing to effectively enhance data complexity and impacting the model's generalization ability. Furthermore, they lack effective collection of VPN establishment and connection error logs, hindering the full utilization of this log data for model training and parameter adjustment, as well as automatic VPN parameter adjustment and connection negotiation based on error logs. This necessitates complex manual configuration and debugging by the user. Existing technologies also fail to collect and accumulate a knowledge base, hindering further improvements in VPN connection accuracy. In summary, existing technologies suffer from complex VPN connection configuration, numerous parameters, difficult debugging, inability to automatically adjust parameters and negotiate connections based on error logs, and an inability to proactively accumulate a knowledge base. Summary of the Invention

[0006] To address at least one of the technical problems mentioned in the background, this invention provides a VPN fast self-negotiation connection method and system. This method simultaneously inputs VPN connection error logs to a multi-threaded retrieval block and a fine-tuned LLM large model module, and finally obtains the final solution based on the evaluation results of both. This allows users to automatically connect to the VPN and adjust parameters automatically based on error logs by inputting only a few basic parameters, thereby reducing the complexity of VPN configuration and improving the convenience and reliability of the connection.

[0007] To achieve the above objectives, the present invention adopts the following technical solution: A first aspect of the present invention provides a VPN fast auto-negotiation connection method, comprising the following steps: Determine if the VPN to be detected has an abnormal connection. If so, call the constructed local knowledge base, which includes an encoding dictionary and a RAG vector knowledge base. Perform multi-threaded retrieval using the encoding dictionary to obtain a first solution. Combine the RAG vector knowledge base and the trained LLM large model to obtain a second solution. Evaluate the first and second solutions to obtain the final solution. Determine if the VPN connection remains abnormal after adjustments based on the final solution. If so, input the previously output second solution and final solution as prompt words back into the trained LLM model until a solution that enables VPN connection is obtained. Update the VPN anomaly log and corresponding solution to the local knowledge base.

[0008] Furthermore, the local knowledge base is built upon historical VPN automatic connection error log data and corresponding solutions, including: The connection error log data and corresponding solutions are organized according to the specified data format to obtain the organized data; The processed data is cleaned, the dictionary is initialized, and the initialized dictionary is stored in the encoding dictionary library. The processed data, along with the merged data from multiple error logs, is vectorized and encoded, and the encoded data is stored in the RAG vector knowledge base.

[0009] Furthermore, the first solution obtained by combining the encoded dictionary library for multi-threaded retrieval includes: starting... n Each thread searches the content in the encoding dictionary. Based on the character length L specified in the dictionary and the number of threads, the thread number is determined. i The start and end positions retrieved by the nth thread, for the nth thread i Each thread has a starting position for its retrieval. The end position of the search is .

[0010] Furthermore, the trained LLM large model is obtained by loading a tokenizer on the server, fine-tuning the LoRa parameters and training epochs and loss, loading weights for LLM model inference training, and adjusting hyperparameters on the test set.

[0011] Furthermore, the evaluation of the first and second solutions to obtain the final solution includes: If the first and second solutions are completely identical, then the same solution will be used as the final solution. If the first solution and the second solution are inconsistent, the first solution shall be taken as the final solution; If the first solution has no content, then the second solution will be used as the final solution.

[0012] Furthermore, updating the VPN anomaly logs and corresponding solutions to the local knowledge base includes embedding the VPN anomaly logs and solutions into a model via embedding, storing the new knowledge in the local RAG vector knowledge base, and cleaning and updating the VPN anomaly logs and solutions to the local encoding dictionary.

[0013] A second aspect of the present invention provides a VPN fast auto-negotiation connection system, comprising: An anomaly detection module is used to determine whether the VPN to be detected has an abnormal connection. If so, it calls the constructed local knowledge base, which includes an encoding dictionary and a RAG vector knowledge base. The solution generation module is used to obtain a first solution by combining the encoding dictionary library for multi-threaded retrieval, and a second solution by combining the RAG vector knowledge base and the trained LLM large model. The first and second solutions are evaluated to obtain the final solution. The solution update module is used to determine whether the VPN connection is still abnormal after being adjusted according to the final solution. If so, the second solution and the final solution output last time are used as prompt words to be input again into the trained LLM model until a solution that makes the VPN connection normal is obtained. The VPN abnormality log and the corresponding solution are then updated to the local knowledge base.

[0014] A third aspect of the present invention provides a computer-readable storage medium.

[0015] A computer-readable storage medium having a computer program stored thereon that, when executed by a processor, implements the steps of a VPN fast auto-negotiation connection method as described above.

[0016] A fourth aspect of the present invention provides a computer device.

[0017] A computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the program, implements the steps of a VPN fast auto-negotiation connection method as described above.

[0018] A fourth aspect of the present invention provides a program product.

[0019] A program product, which is a computer program product, includes a computer program that, when executed by a processor, implements the steps of the VPN fast auto-negotiation connection method described above.

[0020] Compared with the prior art, the beneficial effects of the present invention are: 1. This invention effectively collects VPN establishment logs and connection error logs, and provides data augmentation for VPN connection scenarios. The trained knowledge base and model are downloaded to the device, and the VPN can automatically connect by the user with just a few basic parameters. This eliminates the need for complicated manual configuration, greatly reducing the difficulty of VPN connection configuration and improving the convenience of connection.

[0021] 2. This invention simultaneously inputs the logs of VPN connection anomalies into a multi-threaded retrieval program block and a fine-tuned LLM large model module, and finally outputs corresponding solutions based on both. It automatically adjusts VPN parameters and negotiates connections based on error logs; it also autonomously and continuously accumulates a knowledge base to continuously improve the accuracy of VPN connections.

[0022] Advantages of additional aspects of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0023] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0024] Figure 1 This is an overall flowchart of a VPN fast auto-negotiation connection method provided in an embodiment of the present invention; Figure 2 This is a data flow diagram of a VPN fast auto-negotiation connection method provided in an embodiment of the present invention; Figure 3 This is a schematic diagram of a VPN fast self-negotiation connection device provided in an embodiment of the present invention. Detailed Implementation

[0025] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0026] It should be noted that the following detailed description is illustrative and intended to provide further explanation of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0027] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of exemplary embodiments according to the invention. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0028] Current technologies still rely on remotely transmitting logs to a large server model, using the network for diagnosis and evaluation. They lack data augmentation methods specifically for VPN connection scenarios, failing to effectively enhance data complexity and impacting the model's generalization ability. Furthermore, current technologies lack effective collection of VPN establishment and connection error logs, hindering the full utilization of this log data for model training and parameter tuning. They also cannot automatically adjust VPN parameters based on error logs for automatic connection negotiation, requiring users to perform complex manual configuration and debugging. Finally, they cannot collect and accumulate a knowledge base, further hindering the improvement of VPN connection accuracy. In summary, current technologies suffer from complex VPN connection configurations, numerous parameters, difficult debugging, inability to automatically adjust parameters and negotiate connections based on error logs, and a lack of proactive knowledge base accumulation.

[0029] This invention proposes a method for fast self-negotiation VPN connection based on multi-threaded retrieval and large-model inference. By simultaneously inputting VPN connection error logs into a multi-threaded retrieval block and a fine-tuned LLM large-model module, and finally outputting evaluation code based on both, the final option parameters are provided. This allows users to input only a few basic parameters, and the VPN can automatically connect and adjust parameters based on error logs, thereby reducing the complexity of VPN configuration and improving the convenience and reliability of the connection.

[0030] Example 1 like Figure 1 and Figure 2 As shown, this embodiment provides a VPN fast auto-negotiation connection method, including the following steps: Step 1: Obtain connection error log data and corresponding solutions when the VPN automatically connects; In this embodiment, after obtaining historical VPN automatic connection error log data and corresponding solutions, the data is organized and formatted as follows: { "instruction": "Error connection log A, "input": "", "output": "Change option to optionA" }, { "instruction": "packet from 27.154.56.18:500: initial Main Modemessage received on 58.23.17.102:500 but no connection has been authorizedwith policy=PSK", "input": "", "output": "Change option to optionB; change optionX to optionY" }, { "instruction":"vpn error c", "input": "", "output": "Change option to optionC" } Step 1 of this invention can be used to obtain connection error log data and corresponding solutions during VPN automatic connection to build a local knowledge base.

[0031] Step 2: Based on the acquired connection error log data and corresponding solutions, construct an encoding dictionary and a RAG vector knowledge base; Specifically, the steps include the following: Step 201: Construct an encoding dictionary library based on the acquired connection error log data and corresponding solutions; The instruction data is simplified and cleaned, the instruction-output dictionary is initialized, the dictionary is downloaded to the device, and stored in the local knowledge base - "encoded dictionary library".

[0032] For example, for instruction: "packet from 27.154.56.18:500: initial Main Mode message received on 58.23.17.102:500 but no connection has been authorized with policy=PSK", remove the IP and punctuation, and only extract "packet from initial Main Mode message received on but no connection has been authorized with policy=PSK".

[0033] Step 202: Vectorize the instruction data and the data obtained by merging multiple error logs to construct a RAG vector knowledge base; In this embodiment, the instruction data and the data obtained by merging multiple error logs are vectorized and encoded, and each JSON-encoded input text and output text is encoded and returned as an encoded dictionary.

[0034] In this embodiment, the data fusion of multiple error logs specifically refers to the fusion of JSON format data, such as adding multiple data together. This can increase the amount of data and enhance complexity. This step is mainly to generate large model fine-tuning data.

[0035] In this embodiment, the instruction data and the data fused from multiple error logs are vectorized and encoded, specifically by formatting the data. Because the data used for LoRa training needs to be formatted and encoded before being input into the model, the input text is encoded as input_ids, and the output text is encoded as labels. The encoded results are both multi-dimensional vectors.

[0036] A tokenizer can be used to encode the input and output texts separately, and the input_ids and labels can be stored in a dictionary.

[0037] Step 3: Determine if the VPN connection is abnormal. If so, perform multi-threaded retrieval using the encoding dictionary to obtain the first solution, and combine the RAG vector knowledge base and LLM large model to obtain the second solution. Evaluate the first and second solutions to obtain the final solution. The multi-threaded retrieval module uses a complete information comparison method, which is highly reliable, but sometimes it may fail to match, resulting in no output. The LLM large model module, on the other hand, has a certain learning ability for unknown knowledge and its generalization ability is much higher than that of the multi-threaded retrieval module. Therefore, this embodiment sets dynamic evaluation rules to output effective solutions. The system determines if the VPN connection is abnormal. If so, it sends the cleaned anomaly information to the multi-threaded retrieval module, and then sends the uncleaned data and the output from the local RAG vector knowledge base to the LLM large model. Finally, the evaluation module evaluates the outputs of the two modules and provides the final option output.

[0038] Specifically, the steps include the following: Step 301: Obtain real-time VPN automatic connection error logs and perform multi-threaded retrieval using an encoding dictionary to obtain the first solution; In this embodiment, startup n Each thread searches the dictionary descriptions in the encoding dictionary database. Based on the character length of the dictionary description and the number of threads, the thread determines the nth thread. i The start and end positions of each thread; Specifically, a multi-threaded retrieval module is established, and the retrieval is started through the multi-threaded retrieval module. n Each thread searches the dictionary description "instruction" and outputs the "output" content that matches the "instruction" content, i.e., the correct option.

[0039] For example, the instruction character length is L = 16, number of threads is n = 4.

[0040] The length of characters processed by each thread is 1 L / n = 16 = 4.

[0041] For the i 100 threads (of which) i Starting from 0), the starting position of the search is The end position of the search is .

[0042] Specifically: The first thread ( i = 0) The starting position of the search is The end position of the search is .

[0043] The second thread ( i = 1) The starting position of the search is The end position of the search is .

[0044] The third thread ( i = 2) The starting position of the search is The end position of the search is .

[0045] The fourth thread ( i = 3) The starting position of the search is The end position of the search is .

[0046] Step 302: Combining the RAG vector knowledge base and the LLM large model, a second solution is obtained; Specifically, the tokenizer is loaded on the server, LoRa parameters and training epochs and loss are fine-tuned, weights are loaded for LLM model inference training, hyperparameters are adjusted on the test set, the trained model is downloaded to the device, and the second solution is obtained by combining the RAG vector knowledge base and the large LLM model.

[0047] Step 303: Evaluate the first and second solutions according to the established evaluation rules to obtain the final solution; Specifically, the multi-threaded search module will be called, and its output will be denoted as multi_threaded_search_output; Call the large LLM model and obtain its output result llm_output; In this embodiment, the evaluation rules are set as follows: If the output of the multi-threaded retrieval module and the LLM large model are completely consistent, then output this result directly; If the multi-threaded retrieval module has output results, but they are inconsistent with the output of the LLM large model, the results of the multi-threaded retrieval module will be output first. If the multi-threaded retrieval module does not output any results, then the results of the LLM large model will be output.

[0048] This process can be represented as: if (output of multi-threaded retrieval module == LLM model output) { Directly output option } else if (output of multi-threaded retrieval module && (output of multi-threaded retrieval module != LLM model output)) { Output multi-threaded retrieval module options } else { Output LLM large model options } By effectively collecting VPN establishment logs and connection error logs, data augmentation is achieved for VPN connection scenarios. The trained knowledge base and model are downloaded to the device, and users only need to input a few basic parameters for the VPN to connect automatically, eliminating the need for complex manual configuration. This significantly reduces the difficulty of configuring VPN connections and improves connection convenience.

[0049] Step 4: Determine if the VPN connection is still abnormal after adjustments based on the final solution. If so, input the previous LLM model output and evaluation results as prompt words back into the LLM model until a solution that makes the VPN connection normal is obtained. Update the local knowledge base with the VPN error log and the corresponding solution.

[0050] If the VPN connection is still abnormal after adjusting the options based on the output of the evaluation module, simply input the output options of the previous LLM model and the output options of the evaluation module as prompt words into the LLM model again, and repeat this step. If the VPN connection is normal, update the VPN error log and solution to the local knowledge base. Specifically, VPN anomaly logs and solutions are embedded into the model via embedding, and the new knowledge is stored in the local RAG vector knowledge base.

[0051] Following the method in step 2, first clean the data using the VPN anomaly logs and solutions, and then update the local encoding dictionary with the anomaly logs and output solutions. Step 5: Synchronize the local knowledge base to the cloud to obtain the cloud knowledge base.

[0052] The local knowledge base will synchronize with and retrieve the cloud-based knowledge base when the device's network connection is stable. This continuous collection and accumulation of knowledge and experience will further improve the effectiveness and accuracy of resolving VPN connection issues.

[0053] Using the above method, when a user initiates a VPN connection, they only need to enter the server IP, port, and key; other complex parameters are no longer required. If the initial VPN connection fails, the VPN parameters are automatically adjusted for automatic negotiation. This solution reduces the complexity of VPN configuration and enables quick automatic negotiation and connection.

[0054] The method of this invention reduces the complexity of VPN configuration and has self-learning capabilities, enabling rapid automatic negotiation and connection.

[0055] Example 2 like Figure 3 As shown, this embodiment provides a VPN fast auto-negotiation connection device, including: An anomaly detection module is used to determine whether the VPN to be detected has an abnormal connection. If so, it calls the constructed local knowledge base, which includes an encoding dictionary and a RAG vector knowledge base. The solution generation module is used to obtain a first solution by combining the encoding dictionary library for multi-threaded retrieval, and a second solution by combining the RAG vector knowledge base and the trained LLM large model. The first and second solutions are evaluated to obtain the final solution. The solution update module is used to determine whether the VPN connection is still abnormal after being adjusted according to the final solution. If so, the second solution and the final solution output last time are used as prompt words to be input again into the trained LLM model until a solution that makes the VPN connection normal is obtained. The VPN abnormality log and the corresponding solution are then updated to the local knowledge base.

[0056] It should be noted that the specific implementation of a VPN fast auto-negotiation connection system in this embodiment of the invention is similar to the specific implementation of a VPN fast auto-negotiation connection method in this embodiment of the invention. Please refer to the description in the method section for details. To reduce redundancy, it will not be repeated here.

[0057] Example 3 This embodiment provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the VPN fast auto-negotiation connection method described above.

[0058] Example 4 This embodiment provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the steps in the VPN fast auto-negotiation connection method described above.

[0059] Example 5 This embodiment provides a program product, which is a computer program product including a computer program. When the computer program is executed by a processor, it implements the steps in the VPN fast auto-negotiation connection method described above.

[0060] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of hardware embodiments, software embodiments, or embodiments combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage and optical storage) containing computer-usable program code.

[0061] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0062] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0063] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0064] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc.

[0065] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A VPN fast auto-negotiation connection method, characterized in that, Includes the following steps: Determine if the VPN to be detected has an abnormal connection. If so, call the constructed local knowledge base, which includes an encoding dictionary and a RAG vector knowledge base. Perform multi-threaded retrieval using the encoding dictionary to obtain a first solution. Combine the RAG vector knowledge base and the trained LLM large model to obtain a second solution. Evaluate the first and second solutions to obtain the final solution. Determine whether the VPN connection is still abnormal after adjustment based on the final solution. If so, input the second solution output by the previous LLM model and the final solution obtained from the evaluation as prompt words into the trained LLM model again, and repeat the iteration until a solution that makes the VPN connection normal is obtained. Update the VPN abnormality log and the corresponding solution to the local knowledge base. The encoding dictionary library is built based on the acquired connection error log data and corresponding solutions; The first solution, obtained by combining a coding dictionary with multi-threaded retrieval, includes: starting... n Each thread searches the content in the encoding dictionary. Based on the character length L specified in the dictionary and the number of threads, the thread number is determined. i The start and end positions retrieved by the nth thread, for the nth thread i Each thread has a starting position for its retrieval. The end position of the search is .

2. The VPN fast auto-negotiation connection method as described in claim 1, characterized in that, The local knowledge base is built upon historical VPN automatic connection error log data and corresponding solutions, including: The connection error log data and corresponding solutions are organized according to the specified data format to obtain the organized data; The processed data is cleaned, the dictionary is initialized, and the initialized dictionary is stored in the encoding dictionary library. The processed data, along with the merged data from multiple error logs, is vectorized and encoded, and the encoded data is stored in the RAG vector knowledge base.

3. The VPN fast auto-negotiation connection method as described in claim 1, characterized in that, The trained LLM large model is obtained by loading a tokenizer on the server, fine-tuning the LoRa parameters and training epochs and loss, loading weights for LLM model inference training, and adjusting hyperparameters on the test set.

4. The VPN fast auto-negotiation connection method as described in claim 1, characterized in that, The evaluation of the first and second solutions to obtain the final solution includes: If the first and second solutions are completely identical, then the same solution will be used as the final solution. If the first solution and the second solution are inconsistent, the first solution shall be taken as the final solution; If the first solution has no content, then the second solution will be used as the final solution.

5. A VPN fast auto-negotiation connection method as described in claim 1, characterized in that, The step of updating the VPN anomaly logs and corresponding solutions to the local knowledge base includes embedding the VPN anomaly logs and solutions into a model using embedding to store the new knowledge in the local RAG vector knowledge base, and cleaning the VPN anomaly logs and solutions before updating them to the local encoding dictionary.

6. A VPN fast auto-negotiation connection device, characterized in that, include: An anomaly detection module is used to determine whether the VPN to be detected has an abnormal connection. If so, it calls the constructed local knowledge base, which includes an encoding dictionary and a RAG vector knowledge base. The solution generation module is used to obtain a first solution by combining the encoding dictionary library for multi-threaded retrieval, and a second solution by combining the RAG vector knowledge base and the trained LLM large model. The first and second solutions are evaluated to obtain the final solution. The solution update module is used to determine whether the VPN connection is still abnormal after being adjusted according to the final solution. If so, the second solution output by the previous LLM model and the final solution obtained from the evaluation are used as prompt words to be input into the trained LLM model again, and the iteration is repeated until a solution that makes the VPN connection normal is obtained. The VPN abnormal log and the corresponding solution are updated to the local knowledge base. The encoding dictionary library is built based on the acquired connection error log data and corresponding solutions; The first solution, obtained by combining a coding dictionary with multi-threaded retrieval, includes: starting... n Each thread searches the content in the encoding dictionary. Based on the character length L specified in the dictionary and the number of threads, the thread number is determined. i The start and end positions retrieved by the nth thread, for the nth thread i Each thread has a starting position for its retrieval. The end position of the search is .

7. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the steps in a VPN fast self-negotiation connection method as described in any one of claims 1-5.

8. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps in a VPN fast self-negotiation connection method as described in any one of claims 1-5.

9. A program product, said program product being a computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps in a VPN fast self-negotiation connection method as described in any one of claims 1-5.

Citation Information

Patent Citations

  • Intelligent question answering method and system based on multi-module collaborative optimization

    CN119557409A

  • Power distribution network first-aid repair operation safety inspection scheme generation method, system and related device

    CN120123483A