基于隐私集合求交的大模型双向隐私授权

杨帅; 韩益亮; 吴旭光; 冯浩康; 王圆圆

doi:10.11959/j.issn.2096-109x.2025062

您当前的位置：

首页 >

文章列表页 >

基于隐私集合求交的大模型双向隐私授权

专题：大模型数据安全与隐私保护 | 更新时间：2025-11-08

- 基于隐私集合求交的大模型双向隐私授权
- Bidirectional privacy authorization for large models based on private set intersection
- 网络与信息安全学报 2025年11卷第5期页码：101-114
- 作者机构：
  
  1.武警工程大学密码工程学院，陕西西安710086
  2.武警工程大学反恐指挥信息工程教育部重点实验室，陕西西安710086
- 作者简介：
  
  [ "杨帅（2002- ），男，甘肃平凉人，武警工程大学硕士生，主要研究方向为隐私保护与密码学。" ]
  [ "韩益亮（1977- ），男，甘肃会宁人，武警工程大学教授，主要研究方向为密码学与机密计算。" ]
  [ "吴旭光（1986- ），男，河南商丘人，武警工程大学讲师，主要研究方向为区块链与网络安全。" ]
  [ "冯浩康（2002- ），男，河北邯郸人，武警工程大学硕士生，主要研究方向为密码学与信息安全。" ]
  [ "王圆圆（2001- ），女，江西上饶人，武警工程大学硕士生，主要研究方向为密码学与数据聚合。" ]
- 基金信息：
  
  陕西省自然科学基础研究计划项目(2025JC-YBMS-664)
- DOI：10.11959/j.issn.2096-109x.2025062
  中图分类号： TP309
- 收稿：2025-04-09，
  
  修回：2025-06-24，
  
  纸质出版：2025-10-25
- 稿件说明：
移动端阅览
杨帅, 韩益亮, 吴旭光, 等. 基于隐私集合求交的大模型双向隐私授权[J]. 网络与信息安全学报, 2025, 11(5): 101-114.

YANG S, HAN Y L, WU X G, et al. Bidirectional privacy authorization for large models based on private set intersection[J]. Chinese Journal of Network and Information Security, 2025, 11(5): 101-114.
杨帅, 韩益亮, 吴旭光, 等. 基于隐私集合求交的大模型双向隐私授权[J]. 网络与信息安全学报, 2025, 11(5): 101-114. DOI： 10.11959/j.issn.2096-109x.2025062.

YANG S, HAN Y L, WU X G, et al. Bidirectional privacy authorization for large models based on private set intersection[J]. Chinese Journal of Network and Information Security, 2025, 11(5): 101-114. DOI： 10.11959/j.issn.2096-109x.2025062.

摘要

针对大语言模型多模态接入场景面临的隐私泄露风险，以及传统基于角色和属性的访问控制模型在用户属性多元、访问规模大时难以实现细粒度隐私管控的问题，提出一种基于布谷鸟混淆布隆过滤器的隐私集合求交协议，并构建支持隐私保护的属性访问控制框架。首先，设计布谷鸟混淆布隆过滤器结构，通过引入多重哈希链机制降低元素插入时的碰撞概率，并采用异或操作对过滤器内元素的存储位置进行混淆处理，进一步增强其抗攻击能力；其次，基于该过滤器构建两方隐私集合求交协议，在标准半诚实模型下实现了单轮通信开销；最后，结合基于属性的访问控制模型，设计具备双向隐私验证能力的访问控制框架，利用隐私集合求交技术实现访问策略与用户属性信息的双向隐私保护。实验结果表明，在属性数量为200、并发访问量为10万的场景下，所提方案的策略验证耗时相较于基于椭圆曲线密码的密文策略属性加密（CP-ABC）方案降低约55%，显著提升了高并发访问场景下的处理效率，为大语言模型多模态接入场景提供了兼顾安全与性能的隐私保护方案。

Abstract

To address the privacy leakage risks in multimodal access scenarios of large language models (LLMs) and the limitations of traditional role-and attribute-based access control models in achieving fine-grained privacy control with diverse user attributes and large-scale access

a private set intersection (PSI) protocol based on cuckoo-garbled Bloom filter (CGBF) was proposed

and an attribute-based access control framework supporting privacy protection was constructed. The CGBF structure was designed: the collision probability during element insertion was reduced by introducing a multiple hash chain mechanism

and the storage positions of elements in the filter were garbled through XOR operations

which further enhanced its anti-attack capability. A two-party PSI protocol was constructed based on this filter

achieving single-round communication overhead under the standard semi-honest model. By integrating with the attribute-based access control model

an access control framework with bidirectional privacy verification capability was designed. And the bidirectional privacy protection of access policies and user attribute information was realized using PSI technology. Experimental results show that in scenarios with 200 attributes and 100

000 concurrent accesses

the policy verification time of the proposed scheme is reduced by approximately 55% compared to the CP-ABC scheme based on elliptic curve cryptography

and processing efficiency in high-concurrency access scenarios is significantly improved. This scheme is demonstrated to provide a privacy protection solution that balances security and performance for multimodal access scenarios of LLMs.

关键词

Keywords

references

DAS B C , AMINI M H , WU Y Z . Security and privacy challenges of large language models: a survey [J ] . ACM Computing Surveys , 2025 , 57 ( 6 ): 1 - 39 .

SANDHU R S , COYNE E J , FEINSTEIN H L , et al . Role-based access control models [J ] . Computer , 1996 , 29 ( 2 ): 38 - 47 .

HU V C , KUHN D R , FERRAIOLO D F , et al . Attribute-based access control [J ] . Computer , 2015 , 48 ( 2 ): 85 - 88 .

SAHAI A , WATERS B . Fuzzy identity-based encryption [C ] // Advances in Cryptology-EUROCRYPT 2005 . Berlin : Springer , 2005 : 457 - 473 .

DONG C Y , CHEN L Q , WEN Z K . When private set intersection meets big data: an efficient and scalable protocol [C ] // Proceedings of the 2013 ACM SIGSAC Conference on Computer & Communications Security - CCS '13 . New York : ACM , 2013 : 789 - 800 .

RINDAL P , SCHOPPMANN P . VOLE-PSI: fast OPRF and circuit-PSI from vector-OLE [C ] // Advances in Cryptology – EUROCRYPT 2021 . Berlin : Springer , 2021 : 901 - 930 .

CHEN Y C , HUANG K C . JEDI: joint and effective privacy preserving outsourced set intersection and data integration protocols [J ] . IEEE Transactions on Information Forensics and Security , 2023 , 18 : 4504 - 4514 .

LIU B , RUAN O , SHI R H , et al . Quantum private set intersection cardinality based on bloom filter [J ] . Scientific Reports , 2021 , 11 : 17332 .

HUANG X , ZHANG W F , ZHANG S B . Quantum multi-party private set intersection using single photons [J ] . Physica A: Statistical Mechanics and Its Applications , 2024 , 649 : 129974 .

YE C Q , LI J , YE T , et al . Quantum scheme for private set intersection and union cardinality based on quantum homomorphic encryption [J ] . arXiv preprint , arXiv. 2412 . 01032 , 2024 .

HOU M , WU Y , ZHANG S B . Quantum private set intersection scheme based on bell states [J ] . Axioms , 2025 , 14 ( 2 ): 120 .

ALI M , MOHAJERI J , SADEGHI M R , et al . Attribute-based fine-grained access control for outscored private set intersection computation [J ] . Information Sciences , 2020 , 536 : 222 - 243 .

CAI R S , CHEN L X , ZHU Y Z . Using private set intersection to achieve privacy-preserving authorization for IoT systems [J ] . Journal of Information Security and Applications , 2024 , 83 : 103759 .

KOLESNIKOV V , KUMARESAN R , ROSULEK M , et al . Efficient batched oblivious PRF with applications to private set intersection [C ] // Proceedings of the 2016 ACM SIGSAC Conference on Computer and Communications Security . New York : ACM , 2016 : 818 - 829 .

GHOSH S , NILGES T . An algebraic approach to maliciously secure private set intersection [M ] // Advances in Cryptology – EUROCRYPT 2019 . Cham : Springer International Publishing , 2019 : 154 - 185 .

PINKAS B , SCHNEIDER T , TKACHENKO O , et al . Efficient circuit-based PSI with linear communication [C ] // Proceedings of the 38th Annual International Conference on the Theory and Applications of Cryptographic Techniques , Darmstadt . Berlin : Springer , 2019 : 122 - 153 .

HEMENWAY-FALK B , NOBLE D , OSTROVSKY R . Private set intersection with linear communication from general assumptions [C ] // Proceedings of the 18th ACM Workshop on Privacy in the Electronic Society . New York : ACM , 2019 : 14 - 25 .

PINKAS B , ROSULEK M , TRIEU N , et al . SpOT-light: lightweight private set intersection from sparse OT extension [M ] // Advances in Cryptology – CRYPTO 2019 . Cham : Springer International Publishing , 2019 : 401 - 431 .

GARIMELLA G , PINKAS B , ROSULEK M , et al . Oblivious key-value stores and amplification for private set intersection [M ] // Advances in Cryptology-CRYPTO 2021 . Cham : Springer International Publishing , 2021 : 395 - 425 .

PINKAS B , ROSULEK M , TRIEU N , et al . PSI from PaXoS: fast, malicious private set intersection [M ] // Advances in Cryptology – EUROCRYPT 2020 . Cham : Springer International Publishing , 2020 : 739 - 767 .

RAGHURAMAN S , RINDAL P . Blazing fast PSI from improved OKVS and subfield VOLE [C ] // Proceedings of the 2022 ACM SIGSAC Conference on Computer and Communications Security . New York : ACM , 2022 : 2505 - 2517 .

ZHONG H , ZHU W L , XU Y , et al . Multi-authority attribute-based encryption access control scheme with policy hidden for cloud storage [J ] . Soft Computing , 2018 , 22 ( 1 ): 243 - 251 .

周艺华 , 扈新宇 , 李美奇 , 等 . 云环境下基于属性策略隐藏的可搜索加密方案 [J ] . 网络与信息安全学报 , 2022 , 8 ( 2 ): 112 - 121 .

ZHOU Y H , HU X Y , LI M Q , et al . Searchable encryption scheme based on attribute policy hiding in a cloud environment [J ] . Chinese Journal of Network and Information Security , 2022 , 8 ( 2 ): 112 - 121 .

SHI Y F , QIU S . Delegated key-policy attribute-based set intersection over outsourced encrypted data sets for CloudIoT [J ] . Security and Communication Networks , 2021 , 1 : 5595243 .

LI J G , ZHANG Y C , NING J T , et al . Attribute based encryption with privacy protection and accountability for CloudIoT [J ] . IEEE Transactions on Cloud Computing , 2022 , 10 ( 2 ): 762 - 773 .

刘霞 , 王馨族 , 张涛 , 等 . 支持访问策略部分隐藏的CP-ABE方案 [J ] . 通信学报 , 2024 , 45 ( 10 ): 180 - 190 .

LIU X , WANG X Z , ZHANG T , et al . CP-ABE scheme supporting partially hidden access policy [J ] . Journal on Communications , 2024 , 45 ( 10 ): 180 - 190 .

DARBY M L , NIKOLAOU M . MPC: Current practice and challenges [J ] . Control Engineering Practice , 2012 , 20 ( 4 ): 328 - 342 .

SHAMIR A . How to share a secret [J ] . Communications of the ACM , 1979 , 22 ( 11 ): 612 - 613 .

FAN B , ANDERSEN D G , KAMINSKY M , et al . Cuckoo filter: practically better than bloom [C ] // Proceedings of the 10th ACM International on Conference on Emerging Networking Experiments and Technologies . New York : ACM , 2014 : 75 - 88 .

CHENG R , WU K H , SU Y L , et al . An efficient ECC-based CP-ABE scheme for power IoT [J ] . Processes , 2021 , 9 ( 7 ): 1176 .

浏览量

下载量

CSCD

文章被引用时，请邮件提醒。

提交

工具集

关联资源

AutoVulnFix: 大语言模型驱动的智能搜索自适应漏洞修复系统

基于PIR增强的可搜索加密方案研究

面向LBS的ADS-B报文定位数据本地化差分隐私保护方法

FL-IoV：基于联邦学习的综合车联网入侵检测方法

mmDefender: 基于多模态多语言对抗训练的大语言模型增强方法