[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-119932-en":3,"doc-seo-119932-105":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},119932,4810365810221,"Aurora","https://ap-avatar.wpscdn.com/davatar_155a257f0dc6eb9ab79c44ca47cae57d",8,"Research & Report","From Principle to Practice - Vertical Data Minimization for Machine Learning","Training and deploying predictive machine learning models often requires collecting extensive, fine-grained client data, increasing the likelihood that private information can leak after a breach. Data minimization regulations increasingly require collecting only what is adequate, relevant, and necessary for the task, yet deployment-time compliance with data minimization for machine learning has received limited systematic attention. This work introduces a vertical data minimization workflow that limits full-resolution client data during training and deployment.","From Principle to Practice: Vertical Data Minimization for Machine Learning  \narXiv :2311 . 10500v2 [ cs .LG] 22 Nov 2023  \nRobin Staab ETH Zurich, Switzerland [robin.staab@inf.ethz.ch](robin.staab@inf.ethz.ch)  \nNikola Jovanovi ETH Zurich, Switzerland [nikola.jovanovic@inf.ethz.ch](nikola.jovanovic@inf.ethz.ch)  \nAbstract—Aiming to train and deploy predictive models, organizations collect large amounts of detailed client data, risking the exposure of private information in the event of a breach. To mitigate this, policymakers increasingly demand compliance with the data minimization (DM) principle, restricting data collection to only that data which is relevant and necessary for the task. Despite regulatory pressure, the problem of deploying machine learning models that obey DM has so far received little attention. In this work, we address this challenge ina comprehensive manner. We propose a novel vertical DM (vDM) workflow based on data generalization, which by design ensures that no full-resolution client data is collected during training and deployment of models, benefiting client privacy by reducing the attack surface in case of a breach. We formalize and study the corresponding problem of finding generalizations that both maximize data utility and minimize empirical privacy risk, which we quantify by introducing a diverse set of policyaligned adversarial scenarios. Finally, we propose a range of baseline vDM algorithms, as well as Privacy-aware Tree (PAT), an especially effective vDM algorithm that outperforms all baselines across several settings. We plan to release our code asa publicly available library, helping advance the standardization of DM for machine learning. Overall, we believe our work can help lay the foundation for further exploration and adoption of DM principles in real-world applications.  \n1. Introduction  \nAdvances in machine learning (ML) have enabled organizations to automate tasks such as credit risk scoring [1] or fraud detection [2] . As ML models require large amounts of training samples, organizations increasingly collect different types of detailed client data, hoping to improve the models’performance. The deployment of such models, in turn, necessitates the collection of an even larger amount of highlydetailed client data for inference. These developments have led to growing regulatory concerns regarding the effects of large-scale data collection on individuals’ privacy.  \n1.1. Data Minimization  \nIn an attempt to address this issue, several authorities have developed regulations limiting data collection and processing.  \nMislav Balunovi  \nETH Zurich, Switzerland [mislav.balunovic@inf.ethz.ch](mislav.balunovic@inf.ethz.ch)  \nMartin Vechev ETH Zurich, Switzerland martin.vechev@inf.ethz.ch  \nFigure 1: Vertical data minimization (vDM) can greatly reduce the granularity of the data being collected, while not significantly impacting downstream ML models. We give a full overview of the chosen example in Section 8.4 .  \nMost notably, such concerns are an important part of EU’s General Data Protection Regulation (GDPR) [3], California’s Privacy Rights Act (CPRA) [4], and the recent Blueprint fora U.S. AI Bill of Rights [5] . The GDPR, for example, defines data minimization (DM) in Article 5C as the principle of only collecting and using data that is “adequate, relevant and limited to what is necessary in relation to the purposes for which it is processed”. Similarly, the AI Bill of Rights Blueprint dictates “ensuring that data collection conforms to reasonable expectations and that only data strictly necessary for the specific context is collected.” In the context of ML, this implies that the collection of any personal data (e.g., citizenship) has to be justified by the increase in utility of the resulting model. Furthermore, these regulations also apply for model deployment, a setting that many prior approaches cannot feasibly handle (Section 3) .  \nDM in Practice. DM regulations have already had r","cbCaiiSq50ISBYjN","https://ap.wps.com/l/cbCaiiSq50ISBYjN","pdf",4253584,1,20,"English","en",105,"# Introduction\n## Data Minimization\n## Principled Vertical DM for ML","[{\"question\":\"What challenge does the paper address in deploying machine learning under data minimization requirements?\",\"answer\":\"It addresses the difficulty of deploying machine learning models that comply with data minimization, especially during deployment, where prior approaches are not always feasible.\"},{\"question\":\"How does the proposed vertical data minimization (vDM) workflow improve privacy?\",\"answer\":\"vDM is designed so that no full-resolution client data is collected during training and deployment, reducing the attack surface in the event of a breach.\"},{\"question\":\"What is Privacy-aware Tree (PAT), and how does it perform?\",\"answer\":\"PAT is an especially effective vDM algorithm proposed in the paper. It outperforms the baseline vDM algorithms across several evaluated settings.\"}]","From Principle to Practice - Vertical Data Minimization for Machine Learning | PDF",1785727061,50,{"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},"from-principle-to-practice-vertical-data-minimization-for-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/from-principle-to-practice-vertical-data-minimization-for-machine-learning/119932/",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 deploying machine learning under data minimization requirements?","Question",{"text":75,"@type":76},"It addresses the difficulty of deploying machine learning models that comply with data minimization, especially during deployment, where prior approaches are not always feasible.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"How does the proposed vertical data minimization (vDM) workflow improve privacy?",{"text":80,"@type":76},"vDM is designed so that no full-resolution client data is collected during training and deployment, reducing the attack surface in the event of a breach.",{"name":82,"@type":73,"acceptedAnswer":83},"What is Privacy-aware Tree (PAT), and how does it perform?",{"text":84,"@type":76},"PAT is an especially effective vDM algorithm proposed in the paper. It outperforms the baseline vDM algorithms across several evaluated 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},[93,97,101,105,110,114,119,122,126,129,133],{"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":29,"slug":113},6,"Technology","technology",{"id":115,"doc_module":4,"doc_module_name":46,"category_name":116,"show_sort_weight":117,"slug":118},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":46,"category_name":12,"show_sort_weight":120,"slug":121},30,"research-report",{"id":123,"doc_module":4,"doc_module_name":46,"category_name":124,"show_sort_weight":21,"slug":125},9,"Religion & Spirituality","religion-spirituality",{"id":21,"doc_module":4,"doc_module_name":46,"category_name":127,"show_sort_weight":21,"slug":128},"World Cup","world-cup",{"id":130,"doc_module":4,"doc_module_name":46,"category_name":131,"show_sort_weight":130,"slug":132},10,"Lifestyle","lifestyle",{"id":134,"doc_module":4,"doc_module_name":46,"category_name":135,"show_sort_weight":106,"slug":136},19,"General","general"]