人工智能安全专题DOl:10.12379/j.issn.2096-1057.2024.02.03IssueonArtificialIntelligenceSecurity人工智能算力基础设施安全体系架构研究刘永东张瑶王淼（国家工业信息安全发展研究中心北京100040)(liuyongdong@cics-cert.org,cn)ResearchontheSecurityArchitectureofArtificialIntelligenceComputingInfrastructureLiuYongdong,ZhangYao,andWangMiao(ChinaIndustrialControlSystemCyberEmergencyResponseTeam,Beijing100040)AbstractTheartificialintelligencecomputinginfrastructureisacrucialfoundationforthedevelopmentofartificialintelligence.However,duetoitsdiverseattributes,complexnodes,largenumberofusers,andvulnerabilityofartificialintelligenceitself,theconstructionandoperationofartificialintelligencecomputinginfrastructurefaceseveresecuritychallenges.Thisarticleanalyzestheconnotationandsecuritydevelopmentbackgroundofartificialintelligencecomputinginfrastructure,proposesasecurityarchitectureforartificialintelligencecomputinginfrastructurefromthreeaspects:strengtheningitsownsecurity,ensuringoperationalsecurity,andfacilitatingsecuritycompliance.Itputsforwarddevelopmentsuggestionsaimingtoprovidemethodologicalideasforthesecurityconstructionofartificialintelligencecomputinginfrastructure,offerabasisforselectionanduseofsafeartificialintelligencecomputinginfrastructure,andprovidedecision-makingreferenceforthehealthyandsustainabledevelopmentoftheartificialintelligenceindustry.Keywordsartificialintelligence;artificialintelligencecomputingpower;artificialintelligenceinfrastructure;artificialintelligencesecurity;securityarchitecture摘要人工智能算力基础设施是人工智能发展的重要基石，但由于其属性多样、节点复杂、用户数量多以及人工智能自身脆弱性等特性，使得人工智能算力基础设施在建设和运营过程中面临着严峻的安全挑战.分析了人工智能算力基础设施的内涵和安全发展的背景形势，从强化自身安全、保障运行安全、助力安全合规3方面提出了人工智能算力基础设施安全体系架构及发展建议，旨在为人工智能算力基础设施安全建设提供方法和思路，为选择使用安全的人工智能算力基础设施提供判别依据，为人工智能产业健康、持续发展提供决策参考。关键词人...
