[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-120044-en":3,"doc-seo-120044-105":29,"detail-sidebar-cat-0-en-105":90},{"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":20,"is_downloadable":20,"audit_status":20,"page_count":11,"language":21,"language_code":22,"site_id":23,"html_lang":22,"table_of_contents":24,"faqs":25,"seo_title":26,"seo_description":14,"update_tm":27,"read_time":28},120044,5909877438554,"Maeve","https://ap-avatar.wpscdn.com/avatar/5600025385ad2bf12a7?_k=1778553567797529272",8,"Research & Report","Computational Attestations of Polynomial Integrity Towards Verifiable Machine Learning","Machine-learning services outsourced to untrusted providers create a central tension: consumers need confidence in both privacy protection and correctness while limiting verification effort. This work applies a non-interactive, plausibly post-quantum secure, probabilistically checkable argument system to produce integrity proofs for a privacy mechanism applied during training. The approach proves correct training of a differentially private (DP) linear regression on 60,000 samples in 55 minutes, with full computation verification in 47 seconds, and introduces parallel execution for further runtime reductions.","Computational Attestations of Polynomial Integrity Towards Verifiable Machine Learning  \n[Dustin Ray -dustinray@utexas.edu](Dustin Ray -dustinray@utexas.edu1 and Caroline El)[1](Dustin Ray -dustinray@utexas.edu1 and Caroline El)[ and Caroline El](Dustin Ray -dustinray@utexas.edu1 and Caroline El) [Jazmi -eljazmi@utexas.edu](Jazmi -eljazmi@utexas.edu1)[1](Jazmi -eljazmi@utexas.edu1)  \n1 University of Texas at Austin  \nAbstract—Machine-learning systems continue to advance at a rapid pace, demonstrating remarkable utility in various fields and disciplines. As these systems continue to grow in size and complexity, a nascent industry is emerging which aims to bring machine-learning-as-a-service (MLaaS) to market. Outsourcing the operation and training of these systems to powerful hardware carries numerous advantages, but challenges arise when needing to ensure privacy and the correctness of work carried out by a potentially untrusted party. Recent advancements in the discipline of applied zero-knowledge cryptography, and probabilistic proof systems in general, have led to a means of generating proofs of integrity for any computation, which in turn can be efficiently verified by any party, in any place, at any time.  \nIn this work we present the application of a non-interactive, plausibly-post-quantum-secure, probabilistically-checkable argument system utilized as an efficiently verifiable guarantee that a privacy mechanism was irrefutably applied to a machine-learning model during the training process. That is, we prove the correct training of a differentially-private (DP) linear regression over a dataset of 60,000 samples on a single machine in 55 minutes, verifying the entire computation in 47 seconds. To our knowledge, this result represents the fastest known instance in the literature of provable-DP over a dataset of this size. Finally, we show how this task can be run in parallel, leading to further dramatic reductions in prover and verifier runtime complexity. We believe this result constitutes a key stepping-stone towards end-to-end private MLaaS.  \nIndex Terms—Differential Privacy, Machine-Learning, Linear Regression, Zero-Knowledge Cryptography, Probabilistic Checkable Proofs, ZK-STARK.  \nI. INTRODUCTION  \nThe advent of cloud computing and software-as-a-service systems illustrates an unfolding trend in which many organizations have outsourced large aspects of their computing infrastructure to specialized external service providers, who in turn provide streamlined and robust access to networked hardware and software, often at a much lower cost than selfhosting such infrastructure. With the outsourcing of sensitive computing tasks, there arises a need for security infrastructure that allows for a consumer of such services to be confident that their information is protected and handled according to their specific needs.  \nMachine-learning, particularly generative systems, have continued to advance at a rapid pace, demonstrating utility in a wide variety of manners. A Delloitte study [1] found that at least 50% of surveyed organizations planned to use some form of machine-learning system in 2023 . The growing  \nutility of these systems has largely coincided with their everincreasing complexity and dependence on vast troves of data used in the learning process. Training a machine-learning system is a computationally and financially expensive process, which is often conducted using specialized hardware such as graphics processing units (GPUs) . Access to this hardware is offered through a variety of commercial services, and it is the interaction with these services that presents security challenges for anyone who wishes to leverage a machine-learning system using outsourced hardware on sensitive training data.  \nOur penultimate result in this work shows how an MLaaS operator can employ novel cryptographic techniques with minimal hard assumptions to provide statements of computational integrity to a consumer, such that 1.) [2] Th","cbCainEe6UxGaDx7","https://ap.wps.com/l/cbCainEe6UxGaDx7","pdf",251944,1,"English","en",105,"# Introduction\n## Privacy-Preserving Machine-Learning\n# Cryptographic Integrity Attestations","[{\"question\":\"为什么需要可验证的机器学习（Verifiable ML）？\",\"answer\":\"当模型训练被外包给可能不可信的提供方时，消费者需要证明隐私机制确实被正确应用，并确认计算结果的正确性，同时还要降低验证开销。\"},{\"question\":\"文中使用的核心证明系统是什么？\",\"answer\":\"采用一种非交互、可疑似抗后量子安全、概率可检验的论证系统，用于对训练过程中隐私机制的不可否认性（irrefutably）提供可高效验证的完整性证明。\"},{\"question\":\"该方法在什么任务上进行了验证，效果如何？\",\"answer\":\"证明了在单机上对包含 60,000 个样本的数据集进行不同ially-private（DP）的线性回归训练的正确性。训练耗时 55 分钟，而对整个计算的验证仅需 47 秒；并进一步展示了并行运行可显著降低证明方与验证方的运行复杂度。\"}]","Computational Attestations of Polynomial Integrity Towards Verifiable Machine Learning | PDF",1785727862,20,{"code":4,"msg":30,"data":31},"ok",{"site_id":23,"language":22,"slug":32,"title":13,"keywords":33,"description":14,"schema_data":34,"social_meta":85,"head_meta":87,"extra_data":89,"updated_unix":27},"computational-attestations-of-polynomial-integrity-towards-verifiable-machine-learning","",{"@graph":35,"@context":84},[36,53,67],{"@type":37,"itemListElement":38},"BreadcrumbList",[39,43,47,50],{"item":40,"name":41,"@type":42,"position":20},"https://docshare.wps.com","Home","ListItem",{"item":44,"name":45,"@type":42,"position":46},"https://docshare.wps.com/document/","Document",2,{"item":48,"name":12,"@type":42,"position":49},"https://docshare.wps.com/document/research-report/",3,{"item":51,"name":13,"@type":42,"position":52},"https://docshare.wps.com/document/computational-attestations-of-polynomial-integrity-towards-verifiable-machine-learning/120044/",4,{"url":51,"name":13,"@type":54,"author":55,"headline":13,"publisher":57,"fileFormat":60,"inLanguage":22,"description":14,"dateModified":61,"datePublished":61,"encodingFormat":60,"isAccessibleForFree":62,"interactionStatistic":63},"DigitalDocument",{"name":9,"@type":56},"Person",{"url":40,"name":58,"@type":59},"DocShare","Organization","application/pdf","2026-08-03",true,{"@type":64,"interactionType":65,"userInteractionCount":20},"InteractionCounter",{"@type":66},"ViewAction",{"@type":68,"mainEntity":69},"FAQPage",[70,76,80],{"name":71,"@type":72,"acceptedAnswer":73},"为什么需要可验证的机器学习（Verifiable ML）？","Question",{"text":74,"@type":75},"当模型训练被外包给可能不可信的提供方时，消费者需要证明隐私机制确实被正确应用，并确认计算结果的正确性，同时还要降低验证开销。","Answer",{"name":77,"@type":72,"acceptedAnswer":78},"文中使用的核心证明系统是什么？",{"text":79,"@type":75},"采用一种非交互、可疑似抗后量子安全、概率可检验的论证系统，用于对训练过程中隐私机制的不可否认性（irrefutably）提供可高效验证的完整性证明。",{"name":81,"@type":72,"acceptedAnswer":82},"该方法在什么任务上进行了验证，效果如何？",{"text":83,"@type":75},"证明了在单机上对包含 60,000 个样本的数据集进行不同ially-private（DP）的线性回归训练的正确性。训练耗时 55 分钟，而对整个计算的验证仅需 47 秒；并进一步展示了并行运行可显著降低证明方与验证方的运行复杂度。","https://schema.org",{"og:url":51,"og:type":86,"og:title":13,"og:site_name":58,"og:description":14},"article",{"robots":88,"canonical":51},"index,follow",{"doc_id":7,"site_id":23},{"code":4,"msg":5,"data":91},[92,96,100,104,109,114,119,122,126,129,133],{"id":20,"doc_module":4,"doc_module_name":45,"category_name":93,"show_sort_weight":94,"slug":95},"Story & Novel",90,"story-novel",{"id":46,"doc_module":4,"doc_module_name":45,"category_name":97,"show_sort_weight":98,"slug":99},"Literature",80,"literature",{"id":52,"doc_module":4,"doc_module_name":45,"category_name":101,"show_sort_weight":102,"slug":103},"Exam",70,"exam",{"id":105,"doc_module":4,"doc_module_name":45,"category_name":106,"show_sort_weight":107,"slug":108},5,"Comic",60,"comic",{"id":110,"doc_module":4,"doc_module_name":45,"category_name":111,"show_sort_weight":112,"slug":113},6,"Technology",50,"technology",{"id":115,"doc_module":4,"doc_module_name":45,"category_name":116,"show_sort_weight":117,"slug":118},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":45,"category_name":12,"show_sort_weight":120,"slug":121},30,"research-report",{"id":123,"doc_module":4,"doc_module_name":45,"category_name":124,"show_sort_weight":28,"slug":125},9,"Religion & Spirituality","religion-spirituality",{"id":28,"doc_module":4,"doc_module_name":45,"category_name":127,"show_sort_weight":28,"slug":128},"World Cup","world-cup",{"id":130,"doc_module":4,"doc_module_name":45,"category_name":131,"show_sort_weight":130,"slug":132},10,"Lifestyle","lifestyle",{"id":134,"doc_module":4,"doc_module_name":45,"category_name":135,"show_sort_weight":105,"slug":136},19,"General","general"]