[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-123621-en":3,"doc-seo-123621-105":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},123621,687197100911,"Himbo","https://ap-avatar.wpscdn.com/avatar/a000239b6f1da00475?x-image-process=image/resize,m_fixed,w_180,h_180&k=1785132997149421697",8,"Research & Report","Reclaiming Data Agency in the Age of Ubiquitous Machine Learning - Dissertation","This dissertation addresses why data agency remains necessary in an era of ubiquitous machine learning, where privacy protections are increasingly challenged by modern training and deployment practices. It develops new technical approaches organized around three ideas: disruption-based protection against unauthorized deep learning models, tracing-based data provenance using data isotopes, and direct physical-world attack analysis via backdoors. Comprehensive evaluations assess system effectiveness, robustness to adaptive countermeasures, and practical performance across real and service settings.","THE UNIVERSITY OF CHICAGO  \nRECLAIMING DATA AGENCY IN THE AGE OF UBIQUITOUS MACHINE LEARNING  \nA DISSERTATION SUBMITTED TO  \nTHE FACULTY OF THE DIVISION OF THE PHYSICAL SCIENCES  \nIN CANDIDACY FOR THE DEGREE OF  \nDOCTOR OF PHILOSOPHY  \nDEPARTMENT OF COMPUTER SCIENCE  \nBY  \nEMILY JOY WENGER  \nCHICAGO, ILLINOIS  \nJUNE 2023  \nCopyright 􀀍c 2023 by Emily Joy Wenger  \nAll Rights Reserved  \nTo my parents, my teachers from day one, and my unflagging cheerleaders.  \nAnd to Jon, whose love and encouragement make each day bright.  \nTABLE OF CONTENTS  \nLIST OF FIGURES ....................................... viii  \nLIST OF TABLES ....................................... xi  \nABSTRACT ........................................... xiii  \n1 WHY IS DATA AGENCY NECESSARY? ......................... 1  \n1.1 Existing Data Privacy Solutions ........................... 2  \n1.2 Our Solution: Data Agency .............................. 4  \n2 BACKGROUND AND RELATED WORK ........................ 7  \n2.1 Machine Learning (ML) Overview .......................... 7  \n2.1.1 Training Image Classifiers via Supervised Learning ............. 7  \n2.1.2 Other ML Training Settings and Tasks .................... 8  \n2.2 Attacks and Defenses for ML Image Classification Models ............. 9  \n2.2.1 Adversarial Example Attacks and Defenses ................. 10  \n2.2.2 Poisoning Attacks and Defenses ....................... 11  \n3 DATA AGENCY VIA DISRUPTION—FAWKES: PROTECTING PRIVACY AGAINST UNAUTHORIZED DEEP LEARNING MODELS ......................... 13  \n3.1 Introduction ...................................... 13  \n3.2 Background and Related Work ............................ 15  \n3.2.1 Protecting Privacy via Evasion Attacks ................... 15  \n3.2.2 Protecting Privacy via Poisoning Attacks .................. 16  \n3.2.3 Other Related Work ............................. 17  \n3.3 Protecting Privacy via Cloaking ........................... 18  \n3.3.1 Assumptions and Threat Model ....................... 19  \n3.3.2 Overview and Intuition ............................ 20  \n3.3.3 Computing Cloak Perturbations ....................... 21  \n3.3.4 Cloaking Effectiveness & Transferability .................. 23  \n3.4 The Fawkes Image Cloaking System ......................... 25  \n3.5 System Evaluation .................................. 26  \n3.5.1 Experiment Setup ............................... 27  \n3.5.2 User/Tracker Sharing a Feature Extractor .................. 29  \n3.5.3 User/Tracker Using Different Feature Extractors .............. 31  \n3.5.4 Tracker Models Trained from Scratch .................... 32  \n3.6 Image Cloaking in the Wild .............................. 33  \n3.6.1 Experimental Setup .............................. 33  \n3.6.2 Real World Protection Performance ..................... 34  \n3.7 Trackers with Uncloaked Image Access ....................... 36  \n3.7.1 Impact of Uncloaked Images ......................... 36  \n3.7.2 Sybil Accounts ................................ 37  \n3.7.3 Efficacy of Sybil Images ........................... 39  \n3.8 Countermeasures ................................... 40  \n3.8.1 Cloak Disruption ............................... 41  \n3.8.2 Cloak Detection ............................... 43  \n3.9 Discussion and Conclusion .............................. 44  \n4 DATA AGENCY VIA TRACING—DATA ISOTOPES FOR DATA PROVENANCE INDNNS 46  \n4.1 Introduction ...................................... 46  \n4.2 Requirements and Prior Work ............................. 49  \n4.2.1 Defining Requirements ............................ 49  \n4.2.2 Existing Work on ML Data Provenance ................... 51  \n4.3 Data Isotopes for Data Provenance .......................... 53  \n4.3.1 Provenance via Spurious Correlations .................... 53  \n4.3.2 Introducing Data Isotopes .......................... 54  \n4.4 Data Isotopes Methodology .............................. 56  \n4.4.1 Overview ................................... 57  \n4.4.2 Isotope Creation ...............................","cbCaiebFLgOIIvoy","https://ap.wps.com/l/cbCaiebFLgOIIvoy","pdf",7696466,1,189,"English","en",105,"# Why Is Data Agency Necessary?\n## Existing Data Privacy Solutions\n## Our Solution: Data Agency\n# Background and Related Work\n## Machine Learning (ML) Overview\n## Attacks and Defenses for ML Image Classification Models\n# Data Agency via Disruption—Fawkes: Protecting Privacy Against Unauthorized Deep Learning Models\n## Introduction\n## System Evaluation\n## Image Cloaking in the Wild\n## Countermeasures\n# Data Agency via Tracing—Data Isotopes for Data Provenance\n## Data Isotopes for Data Provenance\n## Evaluating Data Isotopes\n## Isotopes in Real-World Settings\n## Robustness to Adaptive Countermeasures\n# Data Agency via Direct Attack—Backdoor Attacks Against Deep Learning Systems in the Physical World","[{\"question\":\"Why does the dissertation argue that data agency is still necessary in the age of ubiquitous machine learning?\",\"answer\":\"It explains that prevailing privacy solutions face increasing pressure as machine learning systems become widespread and operational. The work frames data agency as a necessary capability to control how data is used and protected.\"},{\"question\":\"How does the dissertation protect privacy against unauthorized deep learning models?\",\"answer\":\"It proposes a disruption-based approach called Fawkes, which focuses on protecting privacy through image cloaking and related countermeasures. The dissertation evaluates system effectiveness, transferability, and performance in real-world conditions.\"},{\"question\":\"What are data isotopes, and how are they used for data provenance?\",\"answer\":\"Data isotopes are introduced as marks that enable provenance detection, including scenarios involving spurious correlations and various isotope creation/detection strategies. The dissertation further evaluates robustness under adaptive countermeasures and considers real-world ML and service settings.\"}]","Reclaiming Data Agency in the Age of Ubiquitous Machine Learning - Dissertation | PDF",1785817682,476,{"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},"reclaiming-data-agency-in-the-age-of-ubiquitous-machine-learning-dissertation","",{"@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/reclaiming-data-agency-in-the-age-of-ubiquitous-machine-learning-dissertation/123621/",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","2026-08-04",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 the dissertation argue that data agency is still necessary in the age of ubiquitous machine learning?","Question",{"text":75,"@type":76},"It explains that prevailing privacy solutions face increasing pressure as machine learning systems become widespread and operational. The work frames data agency as a necessary capability to control how data is used and protected.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"How does the dissertation protect privacy against unauthorized deep learning models?",{"text":80,"@type":76},"It proposes a disruption-based approach called Fawkes, which focuses on protecting privacy through image cloaking and related countermeasures. The dissertation evaluates system effectiveness, transferability, and performance in real-world conditions.",{"name":82,"@type":73,"acceptedAnswer":83},"What are data isotopes, and how are they used for data provenance?",{"text":84,"@type":76},"Data isotopes are introduced as marks that enable provenance detection, including scenarios involving spurious correlations and various isotope creation/detection strategies. The dissertation further evaluates robustness under adaptive countermeasures and considers real-world ML and service settings.","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"]