An edge gateway-based full-device ai content rating and parental control method

Abstract

The application discloses a kind of based on edge gateway's full equipment AI content classification and parental control method, belong to network content management and artificial intelligence security technical field.Gateway identifies AI flow in broadband entrance, and all content review is completed locally, and conversation content is not uploaded to cloud by default.For real-time review of large model generation content and jailbreak induced Prompt, filter according to equipment, time period, user classification.New access equipment default highest security level, support biometric identification and voiceprint automatic identification automatic switching user mode, wake up prior voiceprint before executing instruction.Support AI fraud real-time blocking, emergency warning notification and single equipment independent control.The application solves the problems that existing parental control cannot cover full equipment and cannot filter AI-generated content, providing a unified and convenient solution for home AI security.

Description

Technical Field

[0001] This invention relates to the fields of network content management and artificial intelligence security technology, specifically to a method for implementing AI content rating and parental control across all devices at a broadband gateway. Background Technology

[0002] Existing AI content management has the following shortcomings: existing parental control software is all client-side application, requiring separate installation on each device, and cannot cover devices such as TVs, learning machines, and smart speakers; existing content filtering solutions mainly target traditional web pages and cannot identify content generated by large models; existing solutions cannot block jailbreak attacks targeting large models. Summary of the Invention

[0003] The purpose of this invention is to provide a method for managing AI content across all devices based on an edge gateway. This method enables graded filtering and parental control of AI content across all devices within a local area network without requiring software installation on the terminal. All content review and rule enforcement are completed locally on the gateway, and user AI dialogue content is not uploaded to the cloud by default. Attached Figure Description

[0004] Figure 1 This is an overall flowchart of the method of the present invention; Figure 2 A detailed flowchart for content rating and jailbreak blocking. Detailed Implementation

[0005] The present invention will now be described in detail with reference to the accompanying drawings and embodiments: Example 1: AI-powered safety protection for children in the home Parents can set "child-friendly" content levels for their children's tablets and learning machines on a mobile app and limit AI usage to one hour per day. When a child asks a question to the AI, all responses are filtered through a gateway, automatically blocking inappropriate content such as violence and pornography. If a child attempts to use a jailbroken Prompt to manipulate the AI, the request is directly blocked and the parents are notified. All AI usage records are clearly displayed on the parent's app. Newly connected devices automatically apply the highest level of child-friendly content filtering rules without requiring manual configuration from parents. Example 2: AI Fraud Prevention for Elderly Persons The smart speaker used by an elderly family member connects to the AI ​​service through a gateway. The gateway automatically identifies AI-generated face-swapping voice messages, fake prize winnings, and other fraudulent content, immediately interrupts the conversation, sends an emergency warning notification to the linked parent's account, and automatically blocks further interaction. Example 3: Emergency Warning Notification When the gateway detects that a family member has sent a prompt related to self-harm, suicide, or violence threats, it will automatically send an emergency warning notification to the parent's device in addition to blocking the request, so as to ensure timely intervention. Example 4: Automatic Switching of User Mode Based on Biometric Identification In another embodiment, the gateway integrates a fingerprint recognition module. Parents can unlock the device with a single fingerprint and switch to adult mode; when children use the device, it automatically switches to child mode and filters content. The fingerprint template is encrypted and stored locally in a hardware security module, without being uploaded to the cloud, ensuring privacy and security. The system also supports facial recognition and voiceprint recognition as alternative verification methods. When the confidence level of biometric recognition falls below a preset threshold, the highest security level is maintained by default to ensure security. Example 5: Automatic Voiceprint Recognition and Fraud Prevention The gateway supports automatic voiceprint recognition. When a user interacts with the AI ​​via voice, the gateway automatically identifies the current user using a locally running lightweight voiceprint model and automatically switches the corresponding content level and access permissions. Upon detecting a wake word, the gateway prioritizes local voiceprint recognition verification. Only after successful verification is the subsequent voice command executed; otherwise, the current security level is maintained and an operation log is recorded. When the voiceprint recognition confidence level falls below a preset threshold, the system defaults to the highest restriction level to ensure security. If the gateway detects that an elderly person is interacting with AI-generated fraudulent content, it immediately blocks the conversation and sends an emergency alert to the linked parent's account to prevent the elderly person from being scammed.

Claims

1. A method for full-device AI content rating and parental control based on an edge gateway, characterized in that, Includes the following steps: The gateway identifies large-scale model requests and responses from all devices at the broadband entry point, without requiring any management software to be installed on the terminal; non-AI traffic is not parsed or stored, and is directly forwarded at line speed; all content review and rule execution are completed locally on the gateway, and user AI dialogue content is not uploaded to the cloud by default; the gateway has a built-in customizable content hierarchy library, which supports setting corresponding content levels for each device, each time period, and each user to achieve different levels of AI content filtering; Newly connected devices without configured permission levels in the gateway are subject to the highest-level child-level content blocking rules by default. The gateway performs real-time review of the generated content returned by the large model, filtering out content that does not meet the corresponding device level. This content review targets the content generated by the large model and jailbreak inducement prompts sent by users to the large model, which differs from traditional web page or video content filtering. The gateway identifies user-sent prompts in real time, detects jailbreak inducement statements targeting the large model, and blocks jailbreak requests. Parents can remotely view the AI ​​usage records of all devices via a webpage or app, and set limits on AI usage time, the range of available models, and the types of content that should be prohibited.

2. The method according to claim 1, characterized in that, The content levels include four categories: children, teenagers, adults, and seniors, each corresponding to content filtering rules for users of different age groups.

3. The method according to claim 1, characterized in that, It supports AI fraud detection, and can identify fraudulent content such as face-swapped voice messages, fake prize winnings, and impersonation of relatives and friends generated by large models in real time, and issue warnings to users; when fraudulent content is detected, it immediately interrupts the conversation and pushes an emergency notification to the administrator or parents.

4. The method according to claim 1, characterized in that, The gateway automatically generates and stores a child's AI growth report, which includes AI usage time, interest area analysis, and ability growth curve data. All data is stored locally and is not uploaded to the cloud.

5. The method according to claim 1, characterized in that, The content rating library is customizable, and parents can add or delete specific filter keywords and content types.

6. The method according to claim 1, characterized in that, The gateway detects jailbreak prompts based on a locally running lightweight jailbreak detection model, which supports synchronous updates of the jailbreak rule base in the cloud and can complete the detection without uploading user-requested content.

7. The method according to claim 1, characterized in that, The AI ​​usage record includes the time, device, request type, model name, and content security rating for each AI request.

8. The method according to claim 1, characterized in that, The gateway supports setting a unified policy for different device groups, which are divided by MAC address or IP range.

9. The method according to claim 1, characterized in that, When no content level is assigned to the device, the highest restriction level is used by default to ensure security.

10. The method according to claim 1, characterized in that, When a prompt related to self-harm, suicide, or violent threats is detected, an emergency alert notification is automatically sent to the parent's device in addition to blocking the request.

11. The method according to claim 1, characterized in that, The gateway supports setting AI usage duration and daily request limits for each device. Once the limit is exceeded, AI requests are automatically blocked and a notification is sent to the administrator. It also supports setting usage duration, allowed time periods, and content levels independently for each device, with independent rules for different devices that do not affect each other.

12. The method according to claim 1, characterized in that, It supports automatic identification of the current user through biometric recognition, and automatically switches the content filtering rules and usage time limits corresponding to the user's age level, without the need to manually select a user account; the biometric features include at least one of fingerprint, face, and voiceprint, and the biometric feature template is encrypted and stored locally in the hardware security encryption module and is not uploaded to the cloud; when the confidence level of biometric recognition is lower than the preset threshold, the highest security level is maintained by default.

13. The method according to claim 12, characterized in that, The biometrics include voiceprint recognition. After the gateway detects the wake word, it prioritizes local voiceprint recognition verification. Only after the voiceprint verification is successful will subsequent voice commands be executed. If the verification fails, the current security level is maintained and the operation log is recorded. The system automatically identifies the current user's identity through the user's voice. When the voiceprint recognition confidence level is lower than a preset threshold, it switches to the highest restriction level by default. When it is detected that the user is interacting with AI fraudulent content, the dialogue is blocked in real time and an emergency warning is sent to the parents.

14. A device-wide AI content rating and parental control system based on an edge gateway, characterized in that, The deployment forms include standalone hardware, router integration, and deployment as pure software on existing general-purpose computing devices. The system includes a gateway device and a parental control terminal, and is configured to perform the method described in any one of claims 1 to 13.

Images