2022年中国隐私计算行业洞察报告.pdf

2022 PCview?2022?11?研?PCview?2022?务?2022?11?2?Jason Zhou，?当前，数字化浪潮路歌猛进，数据作为数字化时代的“油”成为重要的产资料。伴随络安全法、数据安全法和个信息保护法的相继颁布与实施，中国将开启数据的可信流通时代，多数据的协同计算与调用模式将被重塑。隐私计算作为数据可信流通中的关键能，拥有袤市场空间。数据源、数据使用者、数据交易所、技术服务商等多元化类型的参与者将在不断探索中推进数据可信流通的稳致远发展。目前隐私计算业处于发展初期阶段，难免存在各种业挑战。在数据要素可信流通建设的重要趋势下，业必将从“多歧路，今安在”的当下，迈向“潮平两岸阔，风正帆悬”的未来！3?务?务?务?PCview?务?(2022?务?)?务?“?TOP10?(2022?务?)”?CONTENTS?Part 1:Industry Development Analysis?Part 3:Vendors Research?Part 2:Practice in Subdivided Industries PCview?Section 1?务?研?务?Development Background PCview?2022?务?2022?11?研?务?n 务?GDP?2005?14.2%?2021?39.8%?研?研?研?研?研?n?务?研?来源1?务?2021?研?6.6ZB?29.4%?研?(67ZB)?9.9%?(16ZB)?研?30%?研?研?研?1?研?务?研?(2021)?PCview?6?务?研?1?105?(?80%?20%?务?)?CIO/CTO/?abcde?76.2%12.4%9.5%1.9%0.0%?%?研?研?务?研?务?(N1=105)?务?务?研?研?研?务?研?研?研?PCview?2022?务?2022?11?研?2020?4?研?务?务?模?研?研?研?研?务?/?务?研?2020?2021?研?PCview?1?N=105(?)?(?80%?20%?务?)?CIO/CTO/?研?190%?研?研?90%?研?API?研?研?研?105?CIO?研?导?务?务?研?务?2021年11?1?2021?9?1?2017?6?1?模?2022-2025?7 PCview?2022?务?2022?11?研?研?研?研?研?研?研?PCview?8?研?3?5?70%?80%?务?研?务?研?研?研?告?研?研?务?研?研?“?研?”?1?105?(?80%?20%?务?)?CIO/CTO/?20.7%62.7%10.4%3.6%2.6%?%N1=105?研?研?(?POC)?/?Section 2?Research on Technical Practice PCview?2022?务?2022?11?PCview?2022?务?2022?11?PCview?10n?(Privacy Preserving Computing)?研?研?(MPC)?(FL)?(TEE)?务?务?n?1?2?3?模?&?4?5?匹?匹?6?7?8?9?10?研?务?研?MPC?(?)?研?研?练?(MPC)?研?(TEE)?(FL)/?(SFL)?研?务?务?研?研?研?研?研?务?研?研?研?TEE?研?研?务?模?研?研?研?研?TEE?研?(1/2)PCview?2022?务?2022?11?模?务?模?PCview?11?务?研?研?务?“?”?“?”?务?研?CPU?务?135?(?研?)?235?务?151?(?)84?（2022，务?，?）?(?)?模?&?匹?匹?，%80.3%70.3%62.4%50.6%49.3%47.6%39.5%40.1%49.0%10.9%?1?2023?6?2?务?研?务?(2/2)?know-how?31.5%?模?告?务?know-how?PCview?2022?务?2022?11?研?(MPC?FL?TEE?)?“?”?务?(?)?研?(1/2)?PCview?1?a?b?/?c)?d?模?务?140?(?90?50?)?266?务?153?(?)113?11.3%18.8%15.0%22.6%20.7%37.6%26.3%47.0%30.1%21.1%32.0%18.8%38.0%22.6%26.7%11.7%?/?1?（2022，务?）?研?务?(?MPC+FL?)?务?53%?12312 PCview?2022?务?2022?11?务?78.7%?务?(2/2)?PCview?务?135?(?研?)?235?务?151?(?)84?,78.7%?,12.8%?,7.2%?,1.3%?1（2022，务?）N=235?1?a?务?&?POC?b?务?&?POC?c?POC务?d?务?“?”?“?”?务?务?务(?)?务?研?1265%?务?80%?13 PCview?2022?务?2022?11?匹?GPU?FPGA?PCview?1?务?务?研?务?135?(?研?)?235?务?151?(?)84?14.9%46.8%85.1%53.2%?（2022，务?）114?务?POC?研?务?务?POC务?123?务?45%?25%?PCview?2022?务?2022?11?务?TEE?务?告?务?告?模?1?3?PCview?15?,29.8%?,48.9%?/?,17.0%?,4.3%?1（2022，务?）?1.?务?135?(?研?)?235?务?151?(?)84?2.?务?51?153?1?a?b?告?c?/?d?1?,18.3%2?3?,19.6%1?3?3?5?,62.1%?1?3?（2022，务?）?1?3?研?PCview?2022?务?2022?11?研?匹?匹?研?研?匹?匹?“?”?匹?模?匹?匹?匹?匹?PCview?16?务?135?(?研?)?235?务?151?(?)84?abcde?%?匹?匹?匹?匹?匹?匹?告?匹?匹?匹?匹?匹?匹?(N1=135)17.0%36.2%31.9%10.6%4.3%?1?N=135?匹?匹?（2022，务?）?易难实实现现难难度度低高?务?匹?匹?务?匹?互互通通程程度度?匹?匹?务?名?匹?匹?务?匹?匹?匹?匹?PCview?2022?务?2022?11?研?务?研?“?”?“?”?务?务?研?know-how?匹?匹?研?研?“?”?“?”?务?研?(?)?研?研?(?)?(?“?”务?研)?PCview?17?1?(2022?务?)?务?2?(2022?务?)?务?(2022，务?)?135?(?研?)?235?务?151?(?)84?,16.6%?,37.9%?,33.6%?/?/?,8.1%?,3.8%?1（2022，务?）?,35.1%?,45.7%?,8.6%?/?,10.6%?2（2022，务?）PCview?2022?务?2022?11?研?研?“?”?“?”?务?“?”?“?”?务?务?研?模?PCview?18?1?a?POC?务?b?c?告?d?e)?f?(2022，务?)1?务?97?(?)?务?195?(?)?“?”?2?务?70%?“?(?筛?)”?务“?”?,32.3%?,25.6%?,12.3%?,14.9%?,10.8%?,4.1%?1(2022，务?)?务?1?“?研?”?模?33%?研?匹?(?研?匹?匹?)?2?3N(?)=97N(?)=195 PCview?2022?务?2022?11?PCview?19基于隐私计算，构建多数据的安全流通能，满业务发展对多数据的链接与调用的需求。核目标如何建设隐私计算平台如何进业务实践主要问题实践策略与建议?(2022，务?)?要点1：根据需求明确技术选型策略。如存在多联合建模场景，FL成为必选项。根据加密要求与性能需求来综合选择具体加密算法以及是否需要融可信硬件。要点2：产品功能。数据管理、节点与资源管理、平台管理等成为基础必备功能。需要配置常用基础算法。因目前产品成熟度相对较低，因在场景实践与业务建模过程中，需要不断增强产品能与场景需求的适配性。体系化或是针对具体客户需求的隐私计算实践咨询，请联系报告研究团队。基于场景与业务需求，链接所需数据源。不同业务需要匹配相应数据源来满数字化实践需求，往往数据源的精准性、差异性、价值性成为业务成功的关键。要点3：异构平台的互通能。随着业务实践中所需链接数据种类的增多，将产对异构平台的数据的链接调用需求。因跨平台互联互通是项值得提前考虑的能。基于业务理解与可用数据，展开场景建模。在隐私保护的前提下，通过多数据协同计算展开联合建模，赋能业务数字化发展。场景建模面需要可用的数据源，另面需要对业务场景拥有深刻理解提升模型可用性。除了提供模型外，部分项目会基于业务分润的合作模式，供应商尝试为客户提供业务增长的全流程解决案。?务1?70%?务?确?40%?务?1?N=90Section 3?Industrial Development Status?“?愿?”?PCview?2022?务?2022?11?务?“?”?研?名?n?TEE?n?研?研?研?研?研?研?名?研?研?研?研?21?PCview?研?/?/?研?研?/?研?(2022，?，务?)?/?研&?务?研?匹?PCview?2022?务?2022?11?务?/?研&?n?研?务?(?映?务?研?)?n?/?研&?务?研?研?务?务?愿?22?PCview?/?|?DataTrust?研&?/?/?务?/?务?愿?(2022，?，务?)n?(?)?(?)?n?“?愿?”?“?愿?”?(?)?愿?愿?PCview?2022?务?2022?11?务?2026?184?103.3%2021?80%?研?模?研?告?模?务?研?PCview?23?研?2021?2022?2026?(2021?2026，?，务?)?5.311.128.262.8125.2184.020212022e2023e2024e2025e2026e2021-2026?务?103.3%CAGR2021-2026e?(?，务?)?1.?2.?研?(?研?)?“?”?研?务?研?3.?(?)?务?(?)?研?研?研?务?PCview?2022?务?2022?11?2021?2022?2016?2022?9?务?57?32?研?研?2020?2021?2022?告?务?PCview?24?1?研?务?研?研?2?研?(?)?(2021?2026，?，务?)1?导?2?导?,17?,1Pre-A?,8A?,12A+?,3Pre-B?,2B?,5B+?,3C?,1?,5?(2016-2022?9?，务?)N=572016?2022?9?0.20 0.25 0.56 2.35 4.62 19.24 5.10 2479141742016201720182019202020212022务?1?2(2016?-2022?9?)?2022(19?)PCview?2022?务?2022?11?(1/2)PCview?2022?务?2022?11?PCview?25?1?65%?务?务?务?模?POC务?POC务?模?1?N=135?1Challenge 1?2Challenge 2?3Challenge 3 PCview?2022?务?2022?11?(2/2)PCview?2022?务?2022?11?PCview?26?“?”?研?“?”?4Challenge 4?5Challenge 5?务?“?”?务?研?研?2022?模?Section 4?Industrial Development Trend?PCview?2022?务?2022?11?研?研?研?研?务?know-how?研?(?)?PCview?28?务?（2022，务?）?80%?60%?研?务?1?N=105.Direction 1Direction 2?务?2?研?know-how?研?1?其中技术部受访者占比其中业务部受访者占比 PCview?2022?务?2022?11?研?PCview?29?研?覆?匹?(?匹?匹?)?研?务?/?研&?研?研?匹?务?研?研?研?1?N=105.务?够?,79.0%?,16.2%?,4.8%?“?务?研?“?1?1?务?2?研?务?务?告?研?务?研?研?研?PCview?2022?务?2022?11?研?研?（?）?（?）?研?研?研?研?研?研?PCview?30?研?匹?匹?&?(2030)?450?研?研?研?&?(2030)?1100?1?3?匹?匹?研?研?务?研?研?CONTENTS?Part 1:Industry Development Analysis?Part 3:Vendors Research?Part 2:Practice in Subdivided Industries PCview?Section 1?Research Model?务?研?务?（DTE Matrix）?PCview?2022?务?2022?11?（DTE?Matrix）?PCview?33?DTE?务?90+?务?确?1?1?/?2?Digital Transformation Efficiency Matrix?DTE Matrix?务?务?3?4务?务?3、?务?RPA?4、?研?务?DTE?务?1?务?DTE?2?务?务?务?1?1 PCview?2022?务?2022?11?DTE?Matrix?PCview?341.?务?DTE?2.?务?务?务?3.?DTE?务?90+?务?确?1?1?/?(?，?，?)?2?1?2?1?3?4?5?6?7?8?9?10?2?2-3?3-5?5?,?Section 2?Research Results?DTE Matrix?务?务?PCview?2022?务?2022?11?(2022，?，?)?PCview?36?2?2-3?3-5?5?2022?11?（?1）?务?2?1.?务?映确?确?愿?2.?务?确?匹?(?匹?)?1?务?“?(2022，?，务?)”?务?Interpretation Instructions37?1?30?务?(?/?)?35?(?)68?103?13?(?务?研?)?务?23?(?)70?93?29?(?)?务?39?(?)61?100?19?(?)?务?30?(?)60?90?研?研?研?PCview?2022?务?2022?11?,29.1%?,53.4%?,4.9%?/?,12.6%?1(2022，务?)?务?研?匹?匹?务?模?研?务?(?)?匹?匹?研?务?模?(1/2)?PCview?1?30?务?(?/?)?35?(?)68?103?(?)?研?务?(?)?研务?(?)?(2022，?，务?)38?&?研?匹?匹?2026?80%?(?研?)?研?70%?研?35%?模?PCview?2022?务?2022?11?(2/2)?PCview?务?1?研?研?务?2?研?研?研?3?研?研?4?ID?研?研务?研?DataTrust?DataTrust?模?ISO?ICDE?ACMCCS?DataTrust?DataTrust?研?DataTrust?研?39?DataTrust?研?研?DataTrust?研?研?研?DataTrust?H5?研?2.?2.?3.?,?1.?1.?PCview?2022?务?2022?11?,40.9%?,46.2%?,9.7%?/?,3.2%?1(2022，务?)?务?研?研?研务?研?研?务?务?(?研?)?研?研?研?-?-?(BOT?build-operate-transfer)?PCview?1?14?(?务?研?)?务?23?(?)70?93?(?)?研?务?(?)?研务?(?)?(2022，?，务?)?&?研?研?2023?研?模?研?研?确?2023?40?(?)?60%?研?PCview?2022?务?2022?11?研?务?研?研?务?2021?2022?/?研?TEE?研?/?研?务?研?研?PCview?(2022，?，务?)?1?29?(?)?务?39?(?)61?100?(?)?研?务?(?)?研务?(?)?&?41?,50.0%?,40.0%?,3.0%?/?,7.0%?1(2022，务?)?模?研?2023?2025?研?研?研?研?75%?研?(?)?研?50%?研?研?名?研?研?名?PCview?2022?务?2022?11?告?研?模?1?3?研?研?(?80%?)?(1/2)?PCview?1?研?19?(?)?务?30?(?)60?90?(?)?研?务?(?)?研务?(?)?(2022，?，务?)?&?研?1?3?4275%?35%?模?研?研?,11.2%?,13.5%?,5.6%?,56.2%?,10.1%?,3.4%?1(2022，务?)N(?)=19N(?)=90 PCview?2022?务?2022?11?(2/2)?PCview?(GWAS)?研?研?研?研?研?研?(GWAS)?1?/?研?研?务?研?2?3?IntelSGX?研?务?iPRIVATES?GWAS?SNPs?SNPs?43?研(?研)?GWAS?iPRIVATES?(GWAS)iPRIVATES?GWAS?告?研?(PCA)?(LR)?(LLR)?GWAS?SNPs?iPRIVATES?(Secure Federated Learning,SFL)?务?务?研?研?研?iPRIVATES?GWAS?iPRIVATES?44?“?(2022，?，?)”?，?PCview?PCview?“?(2022，?，务?)”?务?(?)?“?(2022，?，务?)”?1?研?2?(2022，?，务?)?CONTENTS?Part 1:Industry Development Analysis?Part 3:Vendors Research?Part 2:Practice in Subdivided Industries PCview?Section 1?TOP10?Comprehensive Influence Enterprises（Top 10）?研?务?导?PCview?2022?务?2022?11?(1/3)?PCview?n?务?研?n?n?务?47 PCview?2022?务?2022?11?(2/3)?PCview?TOP 10?（2022，?，务?）?/?1234?30%?1.?2.?务?&?40%30%30%?研?30%?20%?20%?研?(?)?45%30%25%?研?40%30%30%?48?(?/?)?40%30%30%?PCview?2022?务?2022?11?(3/3)?PCview?研?TOP 10?2022，?，务?|DataTrust?49Section 2?(?)?Introduction of Typical Technical Service Providers PCview?2022?务?2022?11?DataTrust?SaaS?DataTrust?PCview?51NRA?DataTrust?研?务?50%?BAT?匹?70%?5?研?DataTrust?(MPC)?(FL)?(DP)?(?研?KMS)?DataTrust?(ECS)?研(RDS)?模?SaaS?ID?DataTrust?务?的?研?模?ISO?/?研?-?-?-?-?ID?ID?ID?纸?务?SQL?&?务LSCC?务?LSCC?务?CSCC?-MPC?-FL?-HE?-DP?-PSI PCview?2022?务?2022?11?DataTrust?研?DataTrust?PCview?DataTrust?MPC?FL?TEE?研?研?研?DataTrust?研?务?研?DataTrust?研?POC?DataTrust?研务?研?爽?研?研?务?研?研?52NRADataTrust?CSCC?务?ID?ID?研?研?LSCC?务?LSCC?务?n DataTrust?研?研?研?研?n DataTrust?对?n DataTrust?MaxCompute?Hadoop?研?研?研?PCview?2022?务?2022?11?MPC?FL?TEE?HE?PCview?(?)?模?研?研?研?务?TEE?CPU?研?研?去?53?务?2?研?100?ARM?务?/?研/?/?/?研?研?(?研?,?研务?研?)?XGBOOST?研?空?空?空?TEETEE?n?nID?nn?研?n?n?(?)n?n?空?研?研?研?模?NRA PCview?2022?务?2022?11?研?PCview?映?2014?iDASH?2021?300+SCI?匹?匹?匹?匹?研?研?务?务?/?研?研?务?研?54?务?(?务?10-20?)?研?研?pSCANNER?研?务?导?务?研?务?务?研?务?AI?研?研?NRA PCview?2022?务?2022?11?PCview?(?)?2021?12?研?Docker?空?映?研?研?研?55NRA?研?PSI?PIR?Docker?空?空?-PG200?-PG1000?-PG30002U2U4U200TPS1000TPS3000TPS?+GPU?+FPGA?PCview?2022?务?2022?11?研?研?PCview?务?爽?研?空?研?匹?务?研?研?务?研?56NRA?空?空?空?研?空?研?;?(?)?SSL?研?研?务?研?研?研?研?模?研?研?务?2022?11?PCview?