Multi-Party Private Table Screening for LLM-Driven ODQA: An Enhanced Method with MPC, Publicly Aggregable Audit, and Dynamic Reputation

HU Rui; WU Hao; PAN Yu-xuan; ZHANG Lin; LIU Yu; ZHU Kong-lin

doi:10.12263/DZXB.20250451

您当前的位置：

首页 >

文章列表页 >

Multi-Party Private Table Screening for LLM-Driven ODQA: An Enhanced Method with MPC, Publicly Aggregable Audit, and Dynamic Reputation

Large-Scale Models and the Internet | 更新时间：2025-12-27

- Multi-Party Private Table Screening for LLM-Driven ODQA: An Enhanced Method with MPC, Publicly Aggregable Audit, and Dynamic Reputation
- ACTA ELECTRONICA SINICA Vol. 53, Issue 9, Pages: 3089-3102(2025)
- 作者机构：
  
  北京邮电大学人工智能学院，北京 100876
- 作者简介：
- 基金信息：
  
  National Key Research and Development Program of China(2023YFB2704500)
- DOI：10.12263/DZXB.20250451
  CLC： TP399;
- Received：31 May 2025，
  
  Accepted：24 September 2025，
  
  Published：25 September 2025
- 稿件说明：
移动端阅览
胡睿, 吴昊, 潘宇轩, 等. 面向LLM开放域问答中多方私有表格筛选：一种MPC可公开聚合审计与动态信誉的增强方法[J]. 电子学报, 2025, 53(09): 3089-3102.

HU Rui, WU Hao, PAN Yu-xuan, et al. Multi-Party Private Table Screening for LLM-Driven ODQA: An Enhanced Method with MPC, Publicly Aggregable Audit, and Dynamic Reputation[J]. Acta Electronica Sinica, 2025, 53(09): 3089-3102.
胡睿, 吴昊, 潘宇轩, 等. 面向LLM开放域问答中多方私有表格筛选：一种MPC可公开聚合审计与动态信誉的增强方法[J]. 电子学报, 2025, 53(09): 3089-3102. DOI：10.12263/DZXB.20250451

HU Rui, WU Hao, PAN Yu-xuan, et al. Multi-Party Private Table Screening for LLM-Driven ODQA: An Enhanced Method with MPC, Publicly Aggregable Audit, and Dynamic Reputation[J]. Acta Electronica Sinica, 2025, 53(09): 3089-3102. DOI：10.12263/DZXB.20250451

摘要

大语言模型（Large Language Model，LLM）驱动的开放域问答（Open-Domain Question Answering，ODAQ）系统，如GIST（Generating Identifiers and Selecting chunks for Tables）框架，在处理海量表格数据时展现出巨大潜力，受到了广泛关注.然而，当ODQA系统需要整合多方私有表格数据进行Top-K候选筛选等环节时，传统方法需要访问全部原数据，这在数据隐私、计算透明度及参与方行为可信度方面面临挑战.虽然现有研究采用零知识证明和基于权益的机制实现了公开可验证性，但在大规模场景下生成和验证单个证明的开销过高，而传统的基于权益的机制在公平性和对动态环境的适应性方面也存在局限性.对此，本文基于多方安全计算（Multi-Party Computation，MPC）、可公开聚合审计与动态信誉机制，提出了一种面向LLM开放域问答中多方私有表格筛选的增强方法.将Top-K多方私有表格筛选过程通过MPC完成，以保护多方私有数据隐私.同时，引入高效的聚合审计机制，将零知识证明技术与随机抽样、聚合证明构造、基于时间窗口的批处理和错误定位相结合，确保评分与排序过程的正确性可以被批量、公开验证.基于区块链的动态信誉反馈机制的集成也增强了系统的公平性，并约束了恶意行为.实验评估表明，本文的Top-K候选筛选方法在保证隐私的同时与GIST原有筛选方法在结果上达到0.91的Top-50平均召回率和0.83的平均Jaccard指数，具有高度一致性，不会影响ODQA端到端任务性能.同时，大规模任务下可公开审计的证明和验证效率均得到提升，与单独的证明相比节省了约87%的证明时间.反馈机制的适应性和公平性也得到了增强.

Abstract

Large language model (LLM) driven open-domain question answering (ODQA) systems

exemplified by frameworks like GIST (Generating Identifiers and Selecting chunks for Tables)

have garnered considerable research attention due to their significant potential in processing extensive tabular data. However

when such ODQA systems integrate data from multiple providers for Top-K candidate screening

traditional methods requiring access to raw data encounter substantial challenges concerning data privacy

computational transparency

and participant trustworthiness. While existing research employs zero-knowledge proofs and stake-based mechanisms to achieve public verifiability

the overhead of generating and verifying individual proofs in large-scale scenarios is often prohibitive. Moreover

conventional stake-based mechanisms exhibit limitations in fairness and adaptability within dynamic environments. This paper proposes an enhanced method for multi-party private table screening in LLM-driven ODQA

which integrates multi-party computation (MPC)

a publicly aggregable audit mechanism

and a dynamic reputation system. This study adapt the Top-K multi-party private table screening process using MPC to ensure data privacy. Concurrently

an efficient aggregable audit mechanism is introduced; this mechanism combines zero-knowledge proof techniques with random sampling

aggregate proof construction

time-window-based batching

and error localization

thereby enabling the public and batch-verified correctness of the scoring and ranking process. The integration of a blockchain-based dynamic reputation feedback mechanism further enhances system fairness and constrains malicious behavior. Experimental evaluations demonstrate that our Top-K candidate screening method

while preserving privacy

achieves high consistency with the original GIST screening approach

attaining a Top-50 average recall of 0.91 and an average Jaccard index of 0.83

thus indicating minimal impact on end-to-end ODQA task performance. Furthermore

the efficiency of publicly auditable proof generation and verification for large-scale tasks is significantly improved

saving approximately 87% of proof time compared to individual proofs. The adaptability and fairness of the feedback mechanism are also demonstrably enhanced.

关键词

Keywords

references

JIN N Z , SIEBERT J , LI D F , et al . A survey on table question answering: Recent advances [M ] // Knowledge Graph and Semantic Computing: Knowledge Graph Empowers the Digital Economy . Singapore : Springer Nature Singapore , 2022 : 174 - 186 .

LIANG X Y , HU R , LIU Y , et al . Open-domain question answering over tables with large language models [C ] // Advanced Intelligent Computing Technology and Applications . Singapore : Springer , 2024 : 347 - 358 .

HARMELINK R , JOOSTEN R , TOPAN E , et al . Data: To share or not to share? A Semi-Systematic Literature Review in (rational) data sharing in inter-organizational systems [J ] . Discover Data , 2024 , 2 ( 1 ): 13 .

HERZIG J , MÜLLER T , KRICHENE S , et al . Open domain question answering over tables via dense retrieval [EB/OL ] . ( 2021-06-09 )[ 2025-05-21 ] . https://arXiv.org/abs/2103.12011 https://arXiv.org/abs/2103.12011 .

YANG Y H , WU J , LONG C N , et al . Blockchain-enabled multiparty computation for privacy preserving and public audit in industrial IoT [J ] . IEEE Transactions on Industrial Informatics , 2022 , 18 ( 12 ): 9259 - 9267 .

KELLER M . MP-SPDZ: A versatile framework for multi-party computation [C ] // Proceedings of the 2020 ACM SIGSAC Conference on Computer and Communications Security . New York : ACM , 2020 : 1575 - 1590 .

LAKHANPAL S , KUMAR S V , RAJ V H , et al . Machine learning-enhanced blockchain solutions for sustainable energy consumption and carbon footprint tracking [C ] // 2024 OPJU International Technology Conference (OTCON) on Smart Computing for Innovation and Advancement in Industry 4.0 . Piscataway : IEEE , 2024 : 1 - 6 .

TAMMINA M R , POSINASETTY B , NAIR P S , et al . Machine learning enabled healthcare balancing patient privacy and data utility [C ] // 2024 Ninth International Conference on Science Technology Engineering and Mathematics . Piscataway : IEEE , 2024 : 1 - 6 .

KAIROUZ P , MCMAHAN H B , AVENT B , et al . Advances and open problems in federated learning [J ] . Foundations and Trends in Machine Learning , 2021 , 14 ( 1/2 ): 1 - 210 .

ZIEMS N , YU W H , ZHANG Z H , et al . Large language models are built-in autoregressive search engines [EB/OL ] . ( 2023-05-16 )[ 2025-05-21 ] . https://arXiv.org/abs/2305.09612 https://arXiv.org/abs/2305.09612 .

CHENG Z J , XIE T B , SHI P , et al . Binding language models in symbolic languages [EB/OL ] . ( 2023-03-01 )[ 2025-05-21 ] . https://arXiv.org/abs/2210.02875 https://arXiv.org/abs/2210.02875 .

LIU Y , HASHIMOTO K , ZHOU Y B , et al . Dense hierarchical retrieval for open-domain question answering [EB/OL ] . ( 2021-10-28 )[ 2025-05-21 ] . https://arXiv.org/abs/2110.15439 https://arXiv.org/abs/2110.15439 .

WU Y X , ZHAO Y , HU B T , et al . An efficient memory-augmented transformer for knowledge-intensive NLP tasks [EB/OL ] . ( 2022-10-30 )[ 2025-05-21 ] . https://arXiv.org/abs/2210.16773 https://arXiv.org/abs/2210.16773 .

DUA D , GUPTA S , SINGH S , et al . Successive prompting for decomposing complex questions [EB/OL ] . ( 2022-12-08 )[ 2025-05-21 ] . https://arXiv.org/abs/2212.04092 https://arXiv.org/abs/2212.04092 .

DESHMUKH P , CARTER B . Secure multi-party computation protocols for privacy-preserving data analysis [J ] . International Journal of Recent Advances in Engineering and Technology , 2023 , 12 ( 2 ): 1 - 6 .

LAVIN R , LIU X K , MOHANTY H , et al . A survey on the applications of zero-knowledge proofs [EB/OL ] . ( 2024-08-01 )[ 2025-05-21 ] . https://arXiv.org/abs/2408.00243 https://arXiv.org/abs/2408.00243 .

ZHU R Y , DING C C , HUANG Y . Efficient publicly verifiable 2PC over a blockchain with applications to financially-secure computations [C ] // Proceedings of the 2019 ACM SIGSAC Conference on Computer and Communications Security . New York : ACM , 2019 : 633 - 650 .

CORDI C , FRANK M P , GABERT K , et al . Auditable, available and resilient private computation on the blockchain via MPC [M ] // Cyber Security, Cryptology, and Machine Learning . Cham : Springer International Publishing , 2022 : 281 - 299 .

KANJALKAR S , ZHANG Y , GANDLUR S , et al . Publicly Auditable MPC-as-a-Service with succinct verification and universal setup [C ] // 2021 IEEE European Symposium on Security and Privacy Workshops . Piscataway : IEEE , 2021 : 386 - 411 .

SEO M . Fair and secure multi-party computation with cheater detection [J ] . Cryptography , 2021 , 5 ( 3 ): 19 .

JIANG Y B , ZHOU Y , FENG T . A blockchain-based secure multi-party computation scheme with multi-key fully homomorphic proxy re-encryption [J ] . Information , 2022 , 13 ( 10 ): 481 .

JIN S , LI Y , CHEN X , et al . Blockchain based publicly auditable multi-party computation with cheater detection [C ] // Information and Communications Security . Singapore : Springer , 2023 : 608 - 626 .

PEDERSEN T P . Non-interactive and information-theoretic secure verifiable secret sharing [C ] // Advances in Cryptology-CRYPTO’91 . Berlin : Springer , 1992 : 129 - 140 .

WENG C K , YANG K , KATZ J , et al . Wolverine: Fast, scalable, and communication-efficient zero-knowledge proofs for Boolean and arithmetic circuits [C ] // 2021 IEEE Symposium on Security and Privacy . Piscataway : IEEE , 2021 : 1074 - 1091 .

CHEN W H , CHANG M W , SCHLINGER E , et al . Open question answering over tables and text [EB/OL ] . ( 2021-02-10 )[ 2025-05-21 ] . https://arXiv.org/abs/2010.10439 https://arXiv.org/abs/2010.10439 .

WANG H Z , YANG Y J , WANG E , et al . Bilateral privacy-preserving worker selection in spatial crowdsourcing [J ] . IEEE Transactions on Dependable and Secure Computing , 2023 , 20 ( 3 ): 2533 - 2546 .

Views

下载量

CSCD

Alert me when the article has been cited

提交

Tools

Publicity Resources

LLM-Based Zero-Shot Imputation of Spatiotemporal Data

Blockchain-Based Hierarchical Federated Learning System

Research on Scalability of Blockchain Based on Sharding: A Survey

Blockchain-Enabled Distributed Spectrum Allocation in Battlefield: A Fast Matching Algorithm

Related Author

MEI Ya-xin

QIN Hui-ling

LIANG Yu-zhu

ZHANG Guang-xue

WANG Tian

HU Rong-lei

LIU Si-hui

DUAN Xiao-yi

Related Institution

School of Artificial Intelligence, Beijing Normal University

Institute of Artificial Intelligence and Future Networks, Beijing Normal University

Department of Electronics and Information Engineering, Beijing Electronic Science and Technology Institute

School of Computer Science, Nanjing University of Posts and Telecommunications

School of Mechanical Engineering, Nanjing University of Science and Technology

⁰