[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-124486-en":3,"doc-seo-124486-105":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},124486,13056703020460,"Valentina","https://ap-avatar.wpscdn.com/avatar/be000253dac470eee5d?_k=1778207105932848923",8,"Research & Report","Comparison of Fully Homomorphic Encryption and Garbled Circuits Approaches in Privacy-Preserving Machine Learning","Privacy-Preserving Machine Learning (PPML) enables inference on private inputs without disclosing sensitive data or proprietary models, addressing growing confidentiality concerns in areas such as healthcare, finance, and NLP. This thesis provides a comparative evaluation of Secure Neural Network Inference using Fully Homomorphic Encryption (FHE) and Garbled Circuits (GC). A two-layer neural network is implemented with CKKS in Microsoft SEAL and TinyGarble2.0 in IntelLabs. Both are assessed under a semi-honest threat model, measuring error, latency, memory, and communication cost, revealing a modular trade-off between execution efficiency and non-interactive inference.","COMPARISON OF FULLY HOMOMORPHIC ENCRYPTION AND GARBLED CIRCUITS APPROACHES IN  \nPRIVACY-PRESERVING MACHINE LEARNING  \nKalyan Cheerla  \nThesis Prepared for the Degree of  \nMASTER OF SCIENCE  \nUNIVERSITY OF NORTH TEXAS  \nJuly 2025  \nAPPROVED:  \nLotfi Ben Othmane, Major Professor Kirill Morozov, Co-Major Professor Song Fu, Committee Member  \nGergely Z´aruba, Chair of the Department of Computer Science and Engineering  \nPaul Krueger, Dean of the College of Engineering  \nVictor Prybutok, Dean of the Toulouse Graduate School  \nCheerla, Kalyan. Comparison of Fully Homomorphic Encryption and Garbled Circuits Approaches in Privacy Preserving Machine Learning. Master of Science (Computer Science), July 2025, 74 pp., 6 tables, 12 figures, references, 69 titles.  \nMachine Learning (ML) is making its way into fields such as healthcare, finance, and natural language processing (NLP), and concerns over data privacy and model confidentiality continue to grow. Privacy-Preserving Machine Learning (PPML) addresses this challenge by enabling inference on private data without revealing sensitive inputs or proprietary models. Leveraging Secure Computation techniques from Cryptography, two widely studied approaches in this domain are Fully Homomorphic Encryption (FHE) and Garbled Circuits (GC) . This thesis presents a comparative evaluation ofFHE and GC for secure neural network inference (SNNI). A two-layer neural network (NN) was implemented using the CKKS scheme from the Microsoft SEAL library (FHE) and the TinyGarble2.0 framework (GC) by IntelLabs. Both implementations are evaluated under a semi-honest threat model, measuring inference output error, round-trip time, peak memory usage, communication overhead, and communication rounds. Results reveal a trade-off: modular GC offers faster execution and lower memory consumption, while FHE supports non-interactive inference. The reproducible implementationsaid secure model deployments in real-world ML-as-a-Service (MLaaS) settings.  \nCopyright 2025 by  \nKalyan Cheerla  \nii  \nTABLE OF CONTENTS  \nPage  \nACKNOWLEDGMENTS viii  \nLIST OF TABLES ix  \nLIST OF FIGURES x  \nCHAPTER 1 INTRODUCTION 1  \n1.1. Problem Statement 2  \n1.1.1. Problem Setup 3  \n1.1.2. Key Comparison Goals 3  \n1.1.3. Assumptions 5  \n1.2. Threat Model 6  \n1.3. Purpose of the Study 7  \n1.4. Contributions 8  \n1.5. Thesis Organization 8  \nCHAPTER 2 Preliminaries 10  \n2.1. Secure Neural Network Inference 10  \n2.1.1. Neural Network Structure 10  \n2.1.2. Motivation for Secure Inference 11  \n2.1.3. Definition and Security Objectives 11  \n2.2. Secure Computation Overview 11  \n2.3. Homomorphic Encryption 14  \n2.3.1. Partially Homomorphic Encryption (PHE) 14  \n2.3.2. Fully Homomorphic Encryption (FHE) 14  \n2.3.3. Leveled and Somewhat Homomorphic Encryption (LHE/SHE) 15  \n2.3.4. Polynomial-Based FHE Schemes 15  \n2.3.5. CKKS Scheme 16  \n2.4. Garbled Circuits 16  \n2.5. Fixed-Point and Polynomial Approximations 17  \n2.5.1. Fixed-Point Representation 17  \n2.5.2. Polynomial Approximations for Non-Linear Functions 18  \n2.5.3. Scaling and Rescaling in HE 19  \nCHAPTER 3 Related Works 21  \n3.1. Secure Function Evaluation (SFE) 21  \n3.1.1. SFE in Secure Neural Network Inference 21  \n3.1.2. SFE in Secure Transformer Inference 22  \n3.1.3. Insights from Recent Surveys 22  \n3.1.4. TinyGarble and Framework Contributions 23  \n3.2. Private Function Evaluation (PFE) 23  \n3.2.1. Oblivious Transfer-Based PFE 23  \n3.2.2. Homomorphic Encryption with Garbled Circuits 24  \n3.2.3. Universal Circuits with Garbled Circuits 24  \n3.2.4. Trusted Execution Environments for PFE 25  \n3.2.5. Summary 25  \n3.3. Non-Interactive Private Function Evaluation (PFE) 25  \n3.3.1. Classical NISC Protocols 26  \n3.3.2. Non-Interactive PFE Using Homomorphic Table Lookup 27  \n3.3.3. Relevance to This Thesis 27  \nCHAPTER 4 Design and Implementation 28  \n4.1. System Overview and Design Goals 28  \n4.1.1. Implementation Overview 28  \n4.1.2. Design Goals 29  \n4.1.3. Use Case Context 29  \n4.2. Evaluation","cbCain84gB459JHJ","https://ap.wps.com/l/cbCain84gB459JHJ","pdf",714896,1,85,"English","en",105,"# Chapter 1 Introduction\n## 1.1. Problem Statement\n## 1.2. Threat Model\n## 1.3. Purpose of the Study\n## 1.4. Contributions\n# Chapter 2 Preliminaries\n## 2.1. Secure Neural Network Inference\n## 2.2. Secure Computation Overview\n## 2.3. Homomorphic Encryption\n## 2.4. Garbled Circuits\n# Chapter 3 Related Works\n## 3.1. Secure Function Evaluation (SFE)\n## 3.2. Private Function Evaluation (PFE)\n## 3.3. Non-Interactive Private Function Evaluation (PFE)\n# Chapter 4 Design and Implementation\n## 4.1. System Overview and Design Goals\n## 4.2. Evaluation Metrics\n## 4.3. Interaction Models\n## 4.4. Fully Homomorphic Encryption-Based Implementation\n## 4.5. Garbled Circuits-Based Implementation\n## 4.6. Implementation Considerations\n# Chapter 5 Comparitive Analysis\n## 5.1. Experimental Setup\n## 5.2. Round-Trip Time Analysis\n## 5.3. Memory Consumption Analysis\n## 5.4. Communication Overhead and Interaction Rounds\n## 5.5. Inference Output Deviation Analysis\n## 5.6. Privacy and Scalability Considerations","[{\"question\":\"What is the goal of privacy-preserving machine learning in this thesis?\",\"answer\":\"Enable inference on private data while preventing disclosure of sensitive inputs and proprietary models. The work targets secure neural network inference as the key use case.\"},{\"question\":\"Which two cryptographic approaches are compared, and what implementations are used?\",\"answer\":\"Fully Homomorphic Encryption (FHE) and Garbled Circuits (GC). FHE uses CKKS in Microsoft SEAL, while GC uses TinyGarble2.0 from IntelLabs.\"},{\"question\":\"How are the approaches evaluated under the threat model?\",\"answer\":\"Both implementations are tested under a semi-honest threat model and compared using metrics including inference output error, round-trip time, peak memory usage, and communication overhead with interaction rounds.\"}]","Comparison of Fully Homomorphic Encryption and Garbled Circuits Approaches in Privacy-Preserving Machine Learning | PDF",1785822733,214,{"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},"comparison-of-fully-homomorphic-encryption-and-garbled-circuits-approaches-in-privacy-preserving-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/comparison-of-fully-homomorphic-encryption-and-garbled-circuits-approaches-in-privacy-preserving-machine-learning/124486/",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","2026-08-04",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 is the goal of privacy-preserving machine learning in this thesis?","Question",{"text":75,"@type":76},"Enable inference on private data while preventing disclosure of sensitive inputs and proprietary models. The work targets secure neural network inference as the key use case.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"Which two cryptographic approaches are compared, and what implementations are used?",{"text":80,"@type":76},"Fully Homomorphic Encryption (FHE) and Garbled Circuits (GC). FHE uses CKKS in Microsoft SEAL, while GC uses TinyGarble2.0 from IntelLabs.",{"name":82,"@type":73,"acceptedAnswer":83},"How are the approaches evaluated under the threat model?",{"text":84,"@type":76},"Both implementations are tested under a semi-honest threat model and compared using metrics including inference output error, round-trip time, peak memory usage, and communication overhead with interaction rounds.","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"]