[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-117419-en":3,"doc-seo-117419-105":30,"detail-sidebar-cat-0-en-105":91},{"code":4,"msg":5,"data":6},0,"success",{"doc_id":7,"user_id":8,"nickname":9,"user_avatar":10,"doc_module":4,"category_id":11,"category_name":12,"doc_title":13,"doc_description":14,"doc_content":15,"file_id":16,"file_url":17,"file_type":18,"file_size":19,"view_count":4,"is_deleted":4,"is_public":20,"is_downloadable":20,"audit_status":20,"page_count":21,"language":22,"language_code":23,"site_id":24,"html_lang":23,"table_of_contents":25,"faqs":26,"seo_title":27,"seo_description":14,"update_tm":28,"read_time":29},117419,7971461740909,"Levi","https://ap-avatar.wpscdn.com/davatar_155a257f0dc6eb9ab79c44ca47cae57d",8,"Research & Report","Oracles for Privacy-Preserving Machine Learning","Deploying machine learning models in production can expose sensitive information, including model parameters, which enables attacks that threaten the privacy of training data. This thesis introduces new privacy-preserving primitives and formal definitions focused on protecting both model parameters and the underlying dataset. It provides security definitions, presents a construction scheme for deployment into production, and gives an informal security argument for the proposed approach.","Oracles for Privacy-Preserving Machine Learning  \nA thesis submitted in partial fulfillment of the requirements for the degree of Master of Science at George Mason University  \nBy  \nMinh Quan Do  \nBachelor of Science  \nGeorge Mason University, 2019  \nDirector: Dr. Foteini Baldimtsi, Professor Department of Computer Science  \nFall Semester 2022  \nGeorge Mason University  \nFairfax, VA  \nCopyright © 2022 by Minh Quan Do All Rights Reserved  \nDedication  \nI dedicate this thesis to all of the wonderful teachers who have helped me get to this point.  \nAcknowledgments  \nI would like to thank the following people who made this possible: Dr. Foteini Baldimtsi, Dr. Evgenios Kornaropoulos, and Dr. Giuseppe Ateniese.  \nTable of Contents  \nPage  \nList of Figures ........................................ vii  \nAbstract ........................................... viii  \n1 Introduction ...................................... 1  \n2 Related Work ...................................... 7  \n2.1 Approaches .................................... 7  \n2.1.1 Cryptographic Approaches ....................... 7  \n2.1.2 Perturbation ............................... 10  \n2.1.3 Privacy-Preserving Dimensionality Reduction (PPDR) ........ 12  \n2.1.4 Hardware-based Approaches ....................... 13  \n2.1.5 Distributed Machine Learning Techniques ............... 14  \n2.2 Attacks ...................................... 18  \n2.2.1 Exploratory Attacks ........................... 18  \n2.2.2 Causitive Attacks ............................. 23  \n3 Definitions ....................................... 24  \n3.1 Cryptographic Preliminaries ........................... 24  \n3.2 Machine Learning Preliminaries ......................... 26  \n4 Defining Privacy Preserving Inference ........................ 30  \n4.1 Definition of Privacy Preserving Inference ................... 30  \n4.1.1 Security Properties ............................ 32  \n4.2 Security Model .................................. 32  \n4.2.1 Inference Server Setting ......................... 33  \n5 Construction ...................................... 36  \n5.1 Interactions Between Oracles & Users ..................... 38  \n5.2 PPIS Construction ................................ 45  \n5.3 Security Analysis ................................. 51  \n5.4 Efficiency ..................................... 54  \n5.5 Ways to Implement Oracles ........................... 54  \n6 Discussion and Extensions .............................. 56  \n6.1 Relations to Related Work ............................ 56  \n6.2 Future Research Directions ........................... 57  \n6.2.1 User-Side Adversaries .......................... 57  \n6.2.2 Oracles for Training ........................... 58  \n6.2.3 Relaxing Ownership Requirements ................... 58  \n7 Conclusion ....................................... 66  \nBibliography ......................................... 67  \nList of Figures  \nFigure Page 2.1 Di↵erent configurations for SplitNN ...................... 17  \n2.2 Di↵erent types of Exploratory Attacks ..................... 19  \n2.3 Fingerprint representations ........................... 20  \n2.4 Fingerprint reconstruction ............................ 20  \n2.5 Images produced by model inversion ...................... 21  \n2.6 Images produced by model inversion based on rounding ........... 22  \n5.1 Preprocessor Oracle Interaction ......................... 40  \n5.2 Model Oracle Interaction ............................ 42  \n5.3 Postprocessor Oracle Interaction ........................ 44  \n6.1 All-Internal Setting ................................ 59  \n6.2 Inference Setting Workflow ........................... 59  \n6.3 Training Server Setting Workflow ........................ 60  \n6.4 Training & Inference Server Setting Workflow ................. 60  \n6.5 Collaborative Training Setting Workflow .................... 61  \n6.6 Training Server Setting Workflow ........................ 61  \n6.7 Federated Learning Setting Workflow ...................... 62  \n6.8 M","cbCaibL0LjjbpXai","https://ap.wps.com/l/cbCaibL0LjjbpXai","pdf",3845449,1,83,"English","en",105,"# Introduction\n# Related Work\n## Approaches\n## Attacks\n# Definitions\n# Defining Privacy Preserving Inference\n# Construction\n## Interactions Between Oracles & Users\n## Security Analysis\n## Efficiency\n# Discussion and Extensions\n# Conclusion\n# Bibliography","[{\"question\":\"Why does deploying machine learning models in production risk privacy?\",\"answer\":\"Model deployment can leak information such as model parameters. That leakage enables attacks that may compromise the privacy of the data used for training.\"},{\"question\":\"What does the thesis propose for privacy preservation?\",\"answer\":\"The thesis introduces definitions for new primitives designed to deploy machine learning models into production while guaranteeing privacy for model parameters and the underlying dataset.\"},{\"question\":\"How is security for the proposed deployment scheme handled?\",\"answer\":\"The thesis provides security definitions, proposes a construction scheme for deploying a model into production, and presents an informal argument supporting the security of the scheme.\"}]","Oracles for Privacy-Preserving Machine Learning | PDF",1785675777,209,{"code":4,"msg":31,"data":32},"ok",{"site_id":24,"language":23,"slug":33,"title":13,"keywords":34,"description":14,"schema_data":35,"social_meta":86,"head_meta":88,"extra_data":90,"updated_unix":28},"oracles-for-privacy-preserving-machine-learning","",{"@graph":36,"@context":85},[37,54,68],{"@type":38,"itemListElement":39},"BreadcrumbList",[40,44,48,51],{"item":41,"name":42,"@type":43,"position":20},"https://docshare.wps.com","Home","ListItem",{"item":45,"name":46,"@type":43,"position":47},"https://docshare.wps.com/document/","Document",2,{"item":49,"name":12,"@type":43,"position":50},"https://docshare.wps.com/document/research-report/",3,{"item":52,"name":13,"@type":43,"position":53},"https://docshare.wps.com/document/oracles-for-privacy-preserving-machine-learning/117419/",4,{"url":52,"name":13,"@type":55,"author":56,"headline":13,"publisher":58,"fileFormat":61,"inLanguage":23,"description":14,"dateModified":62,"datePublished":62,"encodingFormat":61,"isAccessibleForFree":63,"interactionStatistic":64},"DigitalDocument",{"name":9,"@type":57},"Person",{"url":41,"name":59,"@type":60},"DocShare","Organization","application/pdf","2026-08-02",true,{"@type":65,"interactionType":66,"userInteractionCount":4},"InteractionCounter",{"@type":67},"ViewAction",{"@type":69,"mainEntity":70},"FAQPage",[71,77,81],{"name":72,"@type":73,"acceptedAnswer":74},"Why does deploying machine learning models in production risk privacy?","Question",{"text":75,"@type":76},"Model deployment can leak information such as model parameters. That leakage enables attacks that may compromise the privacy of the data used for training.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"What does the thesis propose for privacy preservation?",{"text":80,"@type":76},"The thesis introduces definitions for new primitives designed to deploy machine learning models into production while guaranteeing privacy for model parameters and the underlying dataset.",{"name":82,"@type":73,"acceptedAnswer":83},"How is security for the proposed deployment scheme handled?",{"text":84,"@type":76},"The thesis provides security definitions, proposes a construction scheme for deploying a model into production, and presents an informal argument supporting the security of the scheme.","https://schema.org",{"og:url":52,"og:type":87,"og:title":13,"og:site_name":59,"og:description":14},"article",{"robots":89,"canonical":52},"index,follow",{"doc_id":7,"site_id":24},{"code":4,"msg":5,"data":92},[93,97,101,105,110,115,120,123,128,131,135],{"id":20,"doc_module":4,"doc_module_name":46,"category_name":94,"show_sort_weight":95,"slug":96},"Story & Novel",90,"story-novel",{"id":47,"doc_module":4,"doc_module_name":46,"category_name":98,"show_sort_weight":99,"slug":100},"Literature",80,"literature",{"id":53,"doc_module":4,"doc_module_name":46,"category_name":102,"show_sort_weight":103,"slug":104},"Exam",70,"exam",{"id":106,"doc_module":4,"doc_module_name":46,"category_name":107,"show_sort_weight":108,"slug":109},5,"Comic",60,"comic",{"id":111,"doc_module":4,"doc_module_name":46,"category_name":112,"show_sort_weight":113,"slug":114},6,"Technology",50,"technology",{"id":116,"doc_module":4,"doc_module_name":46,"category_name":117,"show_sort_weight":118,"slug":119},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":46,"category_name":12,"show_sort_weight":121,"slug":122},30,"research-report",{"id":124,"doc_module":4,"doc_module_name":46,"category_name":125,"show_sort_weight":126,"slug":127},9,"Religion & Spirituality",20,"religion-spirituality",{"id":126,"doc_module":4,"doc_module_name":46,"category_name":129,"show_sort_weight":126,"slug":130},"World Cup","world-cup",{"id":132,"doc_module":4,"doc_module_name":46,"category_name":133,"show_sort_weight":132,"slug":134},10,"Lifestyle","lifestyle",{"id":136,"doc_module":4,"doc_module_name":46,"category_name":137,"show_sort_weight":106,"slug":138},19,"General","general"]