[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-117486-en":3,"doc-seo-117486-105":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},117486,1099514067415,"Rowan","https://ap-avatar.wpscdn.com/avatar/100002539d78ffe74a7?x-image-process=image/resize,m_fixed,w_180,h_180&k=1779092875211072502",8,"Research & Report","Privacy-preserving machine learning inference and training with Homomorphic Encryption","Executive summary of a master’s thesis on privacy-preserving machine learning for time series prediction using homomorphic encryption. It introduces the PINStack model, which extends PINPOINT by adding a training algorithm so a single fully connected layer can be trained on encrypted data without exposing contents. The document explains homomorphic encryption fundamentals, key constraints of leveled FHE, and the CKKS scheme’s support for approximate floating-point arithmetic and SIMD vector operations, then evaluates performance against state-of-the-art plaintext methods.","Executive Summary of the Thesis  \nPrivacy-preserving machine learning inference and training with Homomorphic Encryption  \nLaurea Magistrale in Computer Science and Engineering-Ingegneria Informatica Author: Giacomo Mosca  \nAdvisor: Prof. Manuel Roveri  \nCo-advisor: Alessandro Falcetta  \nAcademic year: 2021-2022  \n1. Introduction  \nMachine Learning (ML) techniques have proved to be extremely powerful tools capable of autonomously extracting and learning information from large sets of data, with applications ranging from data mining to medicine and finance. ML tools have become even more accessible thanks to Cloud computing and MachineLearning-as-a-Service (MLaaS) solutions, which offer powerful scalable environments for deep learning models at manageable costs. Outsourcing ML computations to third-party providers however raises important privacy concerns when sensitive data needs to be elaborated. Novel privacy-preserving machine learning techniques have recently emerged to address this issue, in particular making use of Homomorphic Encryption (HE) schemes to perform calculations on encrypted data without ever accessing its contents [1] . The PINPOINT family of models [3] offers a successful application of homomorphic techniques to a privacy-preserving deep learning model for time series prediction. PINPOINT can obtain forecasts on encrypted private data with comparable accuracy to other state-ofthe-art privacy-violating solutions. However,  \nbecause of the limitations imposed by HE schemes the model is only able to perform inference on encrypted data.  \nOur research presents the new privacypreserving PINStack model for time series prediction, which extends the previous PINPOINT architecture by implementing a privacy-preserving training algorithm. PINStack uses the Cheon-Kim-Kim-Song (CKKS) homomorphic encryption scheme [2] to successfully train the parameters of a single fully connected layer on encrypted data without breaching its privacy. The performance of the model is tested under settings that model real use-case scenarios and compared with other state-of-the-art time series prediction solutions executed on plain data.  \n2. Background  \nHomomorphic Encryption (HE) is a typeof encryption scheme that supports the computation of specific operations directly over encrypted data without any knowledge of the encrypted information [1] . Under some assumptions, the result of a homomorphic operation between encrypted values or ciphertexts, when decrypted, will be equal to the one between the  \nExecutive summary Giacomo Mosca  \ncorresponding plain values or plaintexts. Most practical implementations of HE schemes fall under the category of Leveled Fully Homomorphic Encryption, which enables calculations on ciphertexts with the following restrictions:  \n– Only homomorphic additions and multiplications are supported.  \n– Only a set number of operations (specifically multiplications) can be performed on the same ciphertext before its information is lost. Indeed, a noise term is embedded into each ciphertext during encryption to make the scheme secure to decryption, but this noise grows with each homomorphic operation until it overwrites the original encrypted value.  \n– Homomorphic operations come at the cost of important computational overheads.  \nThe Cheon-Kim-Kim-Song (CKKS) homomorphic encryption scheme [2] is chosen for our model due to its native support of floating point values. CKKS is a leveled FHE scheme based on the Ring Learning With Errors (RLWE) problem designed for approximate arithmetic on vectors of complex numbers. The scheme operates on the plaintext space of the polynomial ring R = Z [X]/ (XN + 1) and offers operations to homomorphically evaluate additions, multiplications and rotations over ciphertexts. ARescale operation is used to manage the noise term between multiplications, which in turn determines a maximum multiplicative depth for each ciphertext. The plaintext polynomials of the scheme can encode up to N/","cbCaimk5GG5WQmqm","https://ap.wps.com/l/cbCaimk5GG5WQmqm","pdf",1510867,1,43,"English","en",105,"# Introduction\n## Background","[{\"question\":\"Why is privacy a concern when outsourcing machine learning computations to third parties?\",\"answer\":\"Sensitive data processed by external providers can be exposed during computation. The thesis highlights the need for techniques that avoid accessing encrypted contents while still enabling ML tasks.\"},{\"question\":\"What is the main contribution of the proposed PINStack model?\",\"answer\":\"PINStack enables privacy-preserving training for time series prediction by extending the PINPOINT architecture. It trains parameters of a single fully connected layer directly on encrypted data.\"},{\"question\":\"What role does the CKKS homomorphic encryption scheme play in the approach?\",\"answer\":\"CKKS supports approximate arithmetic on encrypted floating-point values and is used to train and run the model. The scheme manages noise growth via rescaling and supports vector operations through batching.\"}]","Privacy-preserving machine learning inference and training with Homomorphic Encryption | PDF",1785676130,108,{"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},"privacy-preserving-machine-learning-inference-and-training-with-homomorphic-encryption","",{"@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/privacy-preserving-machine-learning-inference-and-training-with-homomorphic-encryption/117486/",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","2026-08-02",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 is privacy a concern when outsourcing machine learning computations to third parties?","Question",{"text":75,"@type":76},"Sensitive data processed by external providers can be exposed during computation. The thesis highlights the need for techniques that avoid accessing encrypted contents while still enabling ML tasks.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"What is the main contribution of the proposed PINStack model?",{"text":80,"@type":76},"PINStack enables privacy-preserving training for time series prediction by extending the PINPOINT architecture. It trains parameters of a single fully connected layer directly on encrypted data.",{"name":82,"@type":73,"acceptedAnswer":83},"What role does the CKKS homomorphic encryption scheme play in the approach?",{"text":84,"@type":76},"CKKS supports approximate arithmetic on encrypted floating-point values and is used to train and run the model. The scheme manages noise growth via rescaling and supports vector operations through batching.","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"]