基于密码的大模型安全治理的思考  

Thoughts on Security Governance of Large Language Models Based on Cryptography

作　　者：田爽 董贵山[1] 王从军[1] 王亚茜 张宇[2] TIAN Shuang;DONG Guishan;WANG Congjun;WANG Yaxi;ZHANG Yu(No.30 Institute of CETC,Chengdu Sichuan 610041,China;Unit 63769 of PLA,Xi'an Shaanxi 710101,China)

机构地区：[1]中国电子科技集团公司第三十研究所,四川成都610041 [2]中国人民解放军63769部队,陕西西安710101

出　　处：《信息安全与通信保密》2024年第6期85-93,共9页Information Security and Communications Privacy

摘　　要：随着大模型技术的迅速发展和广泛应用,其安全性问题日益凸显。密码学作为网络空间安全的核心技术之一,在保障大模型系统安全方面发挥着重要作用。围绕大模型技术发展态势、安全政策与标准现状,针对所面临的安全威胁与风险,研究密码技术在大模型系统安全治理中的应用,并提出基于密码的安全治理架构,阐述架构的构建原理和关键组成,为大模型的安全防护提供了新的视角,为后续密码学与人工智能技术的融合研究提供参考。With the rapid development and extensive application of large language models,its security issues are becoming increasingly prominent.As one of the core technologies for cyberspace security,cryptography plays an crucial role in ensuring the security of large language model systems.Focusing on the development trend of large language model technology,the current status of security policies and standards,and in view of the security threats and risks it faces,this paper delves into the application of cryptography in the security governance of large language model systems,proposes a cryptography-based security governance framework,elucidating the construction principles and key components,which provides a novel perspective on the security protection of large language models and offers a reference for the future research on the integration of cryptography and artificial intelligence technology.

关 键 词：大模型安全 密码技术 网络安全 人工智能 

分 类 号：TP309[自动化与计算机技术—计算机系统结构]