Role large model memory data privacy protection and permission hierarchical management method and system

CN121765752BActive Publication Date: 2026-06-09LIANGSHENG DIGITAL CREATIVE DESIGN (HANGZHOU) CO LTD

Patent Information

Authority / Receiving Office: CN · China
Patent Type: Patents(China)
Current Assignee / Owner: LIANGSHENG DIGITAL CREATIVE DESIGN (HANGZHOU) CO LTD
Filing Date: 2026-03-02
Publication Date: 2026-06-09

AI Technical Summary

Technical Problem

Existing character models cannot reasonably remember and utilize user information while ensuring user privacy and data security during memory data processing, leading to privacy leaks and compliance risks, and making it difficult to balance privacy protection and functional experience.

Method used

By performing sensitivity identification on the memory data generated during the interaction between the large role model and the user, sensitivity scores and permission tags are determined. After desensitization and vector transformation, combined with hierarchical storage and noise reduction, access control and privacy protection of the memory data are achieved.

Benefits of technology

While ensuring user privacy and data security, the role-based big data model is allowed to reasonably remember and utilize user information to meet users' intelligent experience needs and comply with privacy compliance requirements, and to avoid unauthorized use or exposure of sensitive information.

✦ Generated by Eureka AI based on patent content.

Smart Images

Figure CN121765752B_ABST

Patent Text Reader

Abstract

The application discloses a role large model memory data privacy protection and permission hierarchical management method and system. The method comprises the following steps: identifying the sensitivity of memory data, determining the sensitivity score, and then determining the sensitivity level label and the permission label, and storing the sensitivity score, the sensitivity level label and the permission label in association, so as to realize the hierarchical processing of the memory data; based on the memory data fusion vector, the fusion vector is output during semantic retrieval, and the privacy of the memory data is protected; when the role large model and the user carry out memory retrieval in the dialogue scene, the permission of the memory data is controlled, so that the sensitive information is not used by the role large model or is not exposed, and the safety of the memory data is ensured; in the model training scene or the statistical analysis scene, random noise is added to the memory data based on the metadata, so that the privacy of the memory data is protected.

Need to check novelty before this filing date? Find Prior Art