[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-120046-en":3,"doc-seo-120046-105":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},120046,5909877438554,"Maeve","https://ap-avatar.wpscdn.com/avatar/5600025385ad2bf12a7?_k=1778553567797529272",8,"Research & Report","Incentives in Private Collaborative Machine Learning - Differential Privacy Incentive Mechanism","Collaborative machine learning trains models using data from multiple parties, yet participation requires credible incentives and must not amplify privacy risks. The work proposes an incentive mechanism based on differential privacy (DP): each party selects its own DP guarantee and perturbs sufficient statistics accordingly. A mediator values the perturbed statistics via Bayesian surprise, creating a privacy–valuation trade-off that discourages overly strong DP choices. Rewards use different posterior samples, preserving fairness while maintaining DP and high similarity to the grand coalition’s posterior.","arXiv :2404 .0 1676v 1 [ cs .LG] 2 Apr 2024  \nIncentives in Private Collaborative Machine Learning  \nRachael Hwee Ling Sim 1 , Yehong Zhang2 , Trong Nghia Hoang3 Xinyi Xu 1 , Bryan Kian Hsiang Low 1 , and Patrick Jaillet4  \n1 Department of Computer Science, National University of Singapore, Republic of Singapore  \n2 Peng Cheng Laboratory, People’s Republic of China  \n3 School of Electrical Engineering and Computer Science, Washington State University, USA  \n4 Dept. of Electrical Engineering and Computer Science, MIT, USA  \n1 {rachaels,xuxinyi,[lowkh}@comp.nus.edu.sg](lowkh}@comp.nus.edu.sg) , [2](2 zhangyh02@pcl.ac.cn)[ zhangyh02@pcl.ac.cn](2 zhangyh02@pcl.ac.cn)[ ](2 zhangyh02@pcl.ac.cn)[3](3 trongnghia.hoang@wsu.edu)[ trongnghia.hoang@wsu.edu](3 trongnghia.hoang@wsu.edu), [4](4 jaillet@mit.edu)[ jaillet@mit.edu](4 jaillet@mit.edu)  \nAbstract  \nCollaborative machine learning involves training models on data from multiple parties but must incentivize their participation. Existing data valuation methods fairly value and reward each party based on shared data or model parameters but neglect the privacy risks involved. To address this, we introduce differential privacy (DP) as an incentive. Each party can select its required DP guarantee and perturb its sufficient statistic (SS) accordingly. The mediator values the perturbed SS by the Bayesian surprise it elicits about the model parameters. As our valuation function enforces a privacy-valuation trade-off, parties are deterred from selecting excessive DP guarantees that reduce the utility of the grand coalition’s model.  \nFinally, the mediator rewards each party with different posterior samples of the model parameters. Such rewards still satisfy existing incentives like fairness but additionally preserve DP and a high similarity to the grand coalition’s posterior.  \nWe empirically demonstrate the effectiveness and practicality of our approach on synthetic and real-world datasets.  \n1 Introduction  \nCollaborative machine learning (ML) seeks to build ML models of higher quality by training on more data owned by multiple parties [47, 62] . For example, a hospital can improve its prediction of disease progression by training on data collected from more and diversified patients from other hospitals [6] . Likewise, a real-estate firm can improve its prediction of demand and price by training on data from others [9] . However, parties have two main concerns that discourage data sharing and participation in collaborative ML: (a) whether they benefit from the collaboration and (b) privacy.  \nConcern (a) arises as each party would expect the significant cost that it incurs to collect and share data (e.g., the risk of losing its competitive edge) to be covered. Some existing works [47, 51], among other data valuation methods, 1 have recognized that parties require incentives to collaborate, such asa guaranteed fair higher reward from contributing more valuable data than the others, an individually rational higher reward from collaboration than in solitude, and a higher total reward (i.e., group welfare) whenever possible. Often, parties share and are rewarded with information (e.g., gradients [58] or parameters [47] of parametric ML models) computed from the shared data. However, these incentive-aware reward schemes expose parties to privacy risks.  \n1Data valuation methods study how much data is worth. As explained in [46], a party’s data is first valued independently using a performance metric (e.g., see Def. 3.1 later) and then relative to the data contributed by others (e.g., Shapley value (Sec. 4)) . The latter value is helpful to (i) model interpretability and (ii) deciding how much to compensate the data owners fairly.  \n37th Conference on Neural Information Processing Systems (NeurIPS 2023) .  \nOn the other hand, some federated learning (FL) works [34] have addressed the privacy concern (b) and satisfied strict data protection laws (e.g., European Union’s General Data Protection Regu","cbCailP1N3Ni44Av","https://ap.wps.com/l/cbCailP1N3Ni44Av","pdf",1683140,1,39,"English","en",105,"# Introduction\n## Collaborative ML and its incentive concerns\n## Privacy limitations and DP in collaborative settings\n## Proposed privacy-preserving incentive scheme","[{\"question\":\"What challenge does the paper address in private collaborative machine learning?\",\"answer\":\"It addresses how to incentivize parties to participate while preventing privacy risks that can arise from incentive-aware reward schemes.\"},{\"question\":\"How does the proposed mechanism use differential privacy (DP) for incentives?\",\"answer\":\"Each party chooses a required DP guarantee and perturbs its sufficient statistics; the mediator then values the perturbed statistics using Bayesian surprise to incorporate a privacy–valuation trade-off.\"},{\"question\":\"Why do the rewards remain consistent with fairness while preserving privacy?\",\"answer\":\"The mediator rewards parties using different posterior samples, maintaining existing incentive properties like fairness while also satisfying DP and keeping rewards similar to the grand coalition’s posterior.\"}]","Incentives in Private Collaborative Machine Learning - Differential Privacy Incentive Mechanism | PDF",1785727875,98,{"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},"incentives-in-private-collaborative-machine-learning-differential-privacy-incentive-mechanism","",{"@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/incentives-in-private-collaborative-machine-learning-differential-privacy-incentive-mechanism/120046/",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","2026-08-03",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},"What challenge does the paper address in private collaborative machine learning?","Question",{"text":75,"@type":76},"It addresses how to incentivize parties to participate while preventing privacy risks that can arise from incentive-aware reward schemes.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"How does the proposed mechanism use differential privacy (DP) for incentives?",{"text":80,"@type":76},"Each party chooses a required DP guarantee and perturbs its sufficient statistics; the mediator then values the perturbed statistics using Bayesian surprise to incorporate a privacy–valuation trade-off.",{"name":82,"@type":73,"acceptedAnswer":83},"Why do the rewards remain consistent with fairness while preserving privacy?",{"text":84,"@type":76},"The mediator rewards parties using different posterior samples, maintaining existing incentive properties like fairness while also satisfying DP and keeping rewards similar to the grand coalition’s posterior.","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"]