[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-127623-en":3,"doc-seo-127623-105":31,"detail-sidebar-cat-0-en-105":92},{"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":20,"is_deleted":4,"is_public":21,"is_downloadable":21,"audit_status":21,"page_count":22,"language":23,"language_code":24,"site_id":25,"html_lang":24,"table_of_contents":26,"faqs":27,"seo_title":28,"seo_description":14,"update_tm":29,"read_time":30},127623,549768064778,"Finn","https://ap-avatar.wpscdn.com/davatar_6f874abed73319feea01a86fa6f0fab8",8,"Research & Report","Privacy Risk Assessment of Emerging Machine Learning Paradigms - Dissertation","Machine learning (ML) has advanced rapidly, but privacy risks arise when collecting and using data to train increasingly capable models. This dissertation assesses privacy leakage in emerging ML paradigms: it studies membership inference risks from semi-supervised learning, introduces a data-augmentation-based membership inference attack tailored to that setting, and quantifies privacy exposure in self-supervised learning via membership and attribute inference perspectives. It also examines GNN training on graphs by proposing an attack that steals a trained graph and evaluating mitigation strategies.","Saarland University  \nDepartment of Computer Science  \nPrivacy Risk Assessment of Emerging Machine  \nLearning Paradigms  \nDissertation  \nzur Erlangung des Grades  \ndes Doktors der Ingenieurwissenschaften  \nder Fakultät für Mathematik und Informatik  \nder Universität des Saarlandes  \nvon  \nXinlei He  \nSaarbrücken, 2023  \nTag des Kolloquiums: 16 August 2023  \nDekan: Prof. Dr. Jürgen Steimle  \nPrüfungsausschuss:  \nVorsitzender: Prof. Dr. Krishna P. Gummadi  \nBerichterstattende: Dr. Yang Zhang  \nProf. Dr. Neil Zhenqiang Gong Prof. Dr. Isabel Valera  \nAkademischer Mitarbeiter: Dr. Zhengyu Zhao  \nZusammenfassung  \nMaschinelles Lernen (ML) hat enorme Fortschritte gemacht, und Daten sind der Schlüsselfaktor, um diese Entwicklung voranzutreiben. Es gibt jedoch zwei große Herausforderungen bei der Erfassung der Daten und deren Handhabung mit ML-Modellen. Erstens kann die Erfassung qualitativ hochwertiger beschrifteter Daten aufgrund der Notwendigkeit umfangreicher menschlicher Anmerkungen schwierig und teuer sein. Zweitens wurden Graphen genutzt, um die komplexe Beziehung zwischen Entitäten, z.  \nB. sozialen Netzwerken oder Molekülstrukturen, zu modellieren. Herkömmliche MLModelle können Diagrammdaten jedoch aufgrund der nichtlinearen und komplexen Natur der Beziehungen zwischen Knoten möglicherweise nicht effektiv handhaben. Um diesen Herausforderungen zu begegnen, wurden jüngste Entwicklungen im halbüberwachten Lernen und im selbstüberwachten Lernen eingeführt, um unbeschriftete Daten für MLAufgaben zu nutzen. Darüber hinaus wurde eine neue Familie von ML-Modellen, bekanntals Graph Neural Networks, vorgeschlagen, um die Herausforderungen im Zusammenhang mit Graphdaten zu bewältigen. Obwohl sie leistungsfähig sind, sollte auch daspotenzielle Datenschutzrisiko berücksichtigt werden, das sich aus diesen Paradigmenergibt.  \nIn dieser Dissertation führen wir die Datenschutzrisikobewertung der aufkommenden Paradigmen des maschinellen Lernens durch. Erstens untersuchen wir die Datenschutzlecks der Mitgliedschaft, die sich aus halbüberwachtem Lernen ergeben. Konkret schlagen wir den ersten auf Datenaugmentation basierenden Mitgliedschafts-Inferenz-Angriff vor, der auf das Trainingsparadigma halbüberwachter Lernmethoden zugeschnitten ist. Zweitens quantifizieren wir das Durchsickern der Privatsphäre des selbstüberwachten Lernens durch die Linse von Mitgliedschafts-Inferenz-Angriffen und Attribut-InferenzAngriffen. Drittens untersuchen wir die Datenschutzauswirkungen des Trainings von GNNs auf Graphen. Insbesondere schlagen wir den ersten Angriff vor, um einen Graphen aus den Ausgaben eines GNN-Modells zu stehlen, das auf dem Graphen trainiert wird. Schließlich untersuchen wir auch mögliche Verteidigungsmechanismen, um diese Angriffe abzuschwächen.  \nAbstract  \nMachine learning (ML) has progressed tremendously, and data is the key factor to drive such development. However, there are two main challenges regarding collecting the data and handling it with ML models. First, the acquisition of high-quality labeled data can be difficult and expensive due to the need for extensive human annotation. Second, to model the complex relationship between entities, e.g., social networks or molecule structures, graphs have been leveraged. However, conventional ML models may not effectively handle graph data due to the non-linear and complex nature of the relationships between nodes. To address these challenges, recent developments in semi-supervised learning and self-supervised learning have been introduced to leverage unlabeled data for ML tasks. In addition, a new family of ML models known as graph neural networks has been proposed to tackle the challenges associated with graph data. Despite being powerful, the potential privacy risk stemming from these paradigms should also be taken into account.  \nIn this dissertation, we perform the privacy risk assessment of the emerging machine learning paradigms. Firstly, we investigate the membership privacy leakage s","cbCaiey8UZk3k4Vz","https://ap.wps.com/l/cbCaiey8UZk3k4Vz","pdf",7667469,2,1,136,"English","en",105,"# Background of this Dissertation\n## Summary of research contributions","[{\"question\":\"What privacy risks does the dissertation focus on in emerging ML paradigms?\",\"answer\":\"It focuses on membership privacy leakage from semi-supervised learning, privacy leakage in self-supervised learning analyzed through membership and attribute inference, and privacy implications of training graph neural networks (GNNs) on graphs.\"},{\"question\":\"What attack is proposed for semi-supervised learning?\",\"answer\":\"The dissertation proposes a data augmentation-based membership inference attack tailored to the training paradigm of semi-supervised learning methods.\"},{\"question\":\"How does the dissertation study privacy risks for graph neural networks (GNNs)?\",\"answer\":\"It proposes an attack to steal a graph from the outputs of a GNN model trained on that graph, and it explores defense mechanisms to mitigate these attacks.\"}]","Privacy Risk Assessment of Emerging Machine Learning Paradigms - Dissertation | PDF",1785940332,343,{"code":4,"msg":32,"data":33},"ok",{"site_id":25,"language":24,"slug":34,"title":13,"keywords":35,"description":14,"schema_data":36,"social_meta":87,"head_meta":89,"extra_data":91,"updated_unix":29},"privacy-risk-assessment-of-emerging-machine-learning-paradigms-dissertation","",{"@graph":37,"@context":86},[38,54,69],{"@type":39,"itemListElement":40},"BreadcrumbList",[41,45,48,51],{"item":42,"name":43,"@type":44,"position":21},"https://docshare.wps.com","Home","ListItem",{"item":46,"name":47,"@type":44,"position":20},"https://docshare.wps.com/document/","Document",{"item":49,"name":12,"@type":44,"position":50},"https://docshare.wps.com/document/research-report/",3,{"item":52,"name":13,"@type":44,"position":53},"https://docshare.wps.com/document/privacy-risk-assessment-of-emerging-machine-learning-paradigms-dissertation/127623/",4,{"url":52,"name":13,"@type":55,"author":56,"headline":13,"publisher":58,"fileFormat":61,"inLanguage":24,"description":14,"dateModified":62,"datePublished":63,"encodingFormat":61,"isAccessibleForFree":64,"interactionStatistic":65},"DigitalDocument",{"name":9,"@type":57},"Person",{"url":42,"name":59,"@type":60},"DocShare","Organization","application/pdf","2026-08-23","2026-08-05",true,{"@type":66,"interactionType":67,"userInteractionCount":20},"InteractionCounter",{"@type":68},"ViewAction",{"@type":70,"mainEntity":71},"FAQPage",[72,78,82],{"name":73,"@type":74,"acceptedAnswer":75},"What privacy risks does the dissertation focus on in emerging ML paradigms?","Question",{"text":76,"@type":77},"It focuses on membership privacy leakage from semi-supervised learning, privacy leakage in self-supervised learning analyzed through membership and attribute inference, and privacy implications of training graph neural networks (GNNs) on graphs.","Answer",{"name":79,"@type":74,"acceptedAnswer":80},"What attack is proposed for semi-supervised learning?",{"text":81,"@type":77},"The dissertation proposes a data augmentation-based membership inference attack tailored to the training paradigm of semi-supervised learning methods.",{"name":83,"@type":74,"acceptedAnswer":84},"How does the dissertation study privacy risks for graph neural networks (GNNs)?",{"text":85,"@type":77},"It proposes an attack to steal a graph from the outputs of a GNN model trained on that graph, and it explores defense mechanisms to mitigate these attacks.","https://schema.org",{"og:url":52,"og:type":88,"og:title":13,"og:site_name":59,"og:description":14},"article",{"robots":90,"canonical":52},"index,follow",{"doc_id":7,"site_id":25},{"code":4,"msg":5,"data":93},[94,98,102,106,111,116,121,124,129,132,136],{"id":21,"doc_module":4,"doc_module_name":47,"category_name":95,"show_sort_weight":96,"slug":97},"Story & Novel",90,"story-novel",{"id":20,"doc_module":4,"doc_module_name":47,"category_name":99,"show_sort_weight":100,"slug":101},"Literature",80,"literature",{"id":53,"doc_module":4,"doc_module_name":47,"category_name":103,"show_sort_weight":104,"slug":105},"Exam",70,"exam",{"id":107,"doc_module":4,"doc_module_name":47,"category_name":108,"show_sort_weight":109,"slug":110},5,"Comic",60,"comic",{"id":112,"doc_module":4,"doc_module_name":47,"category_name":113,"show_sort_weight":114,"slug":115},6,"Technology",50,"technology",{"id":117,"doc_module":4,"doc_module_name":47,"category_name":118,"show_sort_weight":119,"slug":120},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":47,"category_name":12,"show_sort_weight":122,"slug":123},30,"research-report",{"id":125,"doc_module":4,"doc_module_name":47,"category_name":126,"show_sort_weight":127,"slug":128},9,"Religion & Spirituality",20,"religion-spirituality",{"id":127,"doc_module":4,"doc_module_name":47,"category_name":130,"show_sort_weight":127,"slug":131},"World Cup","world-cup",{"id":133,"doc_module":4,"doc_module_name":47,"category_name":134,"show_sort_weight":133,"slug":135},10,"Lifestyle","lifestyle",{"id":137,"doc_module":4,"doc_module_name":47,"category_name":138,"show_sort_weight":107,"slug":139},19,"General","general"]