Protecting Privacy through Homomorphic Encryption

This book summarizes recent inventions, provides guidelines and recommendations, and demonstrates many practical applications of homomorphic encryption. This collection of papers represents the combined wisdom of the community of leading experts on homomorphic encryption. In the past 3 years, a global community consisting of researchers in academia, industry, and government, has been working closely to standardize homomorphic encryption. This is the first publication of whitepapers created by these experts that comprehensively describes the scientific inventions, presents a concrete security analysis, and broadly discusses applicable use scenarios and markets. This book also features a collection of privacy-preserving machine learning applications powered by homomorphic encryption designed by groups of top graduate students worldwide at the Private AI Bootcamp hosted by Microsoft Research. The volume aims to connect non-expert readers with this important new cryptographic technology in an accessible and actionable way. Readers who have heard good things about homomorphic encryption but are not familiar with the details will find this book full of inspiration. Readers who have preconceived biases based on out-of-date knowledge will see the recent progress made by industrial and academic pioneers on optimizing and standardizing this technology. A clear picture of how homomorphic encryption works, how to use it to solve real-world problems, and how to efficiently strengthen privacy protection, will naturally become clear. Preface References Contents Part I Introduction to Homomorphic Encryption Introduction to Homomorphic Encryption and Schemes 1 Introduction to Homomorphic Encryption 1.1 Plaintexts and Operations 1.2 Vectors and Special-Purpose Plaintext Data Types 1.3 Ciphertexts 1.4 Symmetric vs. Public-Key Homomorphic Encryption 1.5 Parameters and Security 2 The BGV and BFV Encryption Schemes 2.1 Homomorphic Operations Two-Argument Operations Unary Operations 2.2 Parameter Selection 2.3 A BGV/BFV Hello World Example 2.4 Further Information Maintenance Operations Evaluation Keys Data Encoding Data Movement Operations References for the BFV Encryption Scheme References for the BGV Encryption Scheme 3 The CKKS Encryption Scheme 3.1 Homomorphic Operations Two-Argument Operations Unary Operations 3.2 Parameter Selection 3.3 A CKKS Hello World Example 3.4 Further Information Data Encoding Maintenance Operations Evaluation Keys References for the CKKS Scheme Reference Implementations 4 The DM (FHEW) and CGGI (TFHE) Schemes 4.1 Basic Concepts 4.2 Homomorphic Operations Simple Mode Plaintext Space and Operations A DM/CGGI Hello World Example (Using Simple Mode) Advanced Mode Plaintext Space and Operations Advanced-Mode CGGI Hello World Example (Corresponds to the DFA in Fig. 1) 4.3 Further Information Advanced Notes on Parameters Some More Advanced Operations Are Supported Maintenance Operations (and More) Advanced Functionality in the CGGI Encryption Scheme Difference Between DM and CGGI Variants of DM/CGGI Scheme Switching Using CGGI Reference Implementations References Part II Homomorphic Encryption Security Standard Homomorphic Encryption Standard 1 Homomorphic Encryption Standard Section 1: Recommended Encryption Schemes 1.1 Notation and Definitions 1.2 Properties 1.3 The BGV and B/FV Homomorphic Encryption Schemes 1.4 The GSW Scheme and Bootstrapping 1.5 Other Schemes 1.6 Additional Features & Discussion 2 Homomorphic Encryption Standard Section 2: Recommended Security Parameters 2.1 Hard Problems 2.2 Attacks on LWE and Their Complexity 2.3 The Arora-Ge Attack 2.4 Algebraic Attacks on Instances of Ring-LWE 2.5 Secure Parameter Selection for Ring LWE Organizers Contributors References Software References for 7 Homomorphic Encryption Libraries Part III Applications of Homomorphic Encryption Privacy-Preserving Data Sharing and Computation Across Multiple Data Providers with Homomorphic Encryption 1 Motivation 2 System Models and Use Cases 3 Stakeholders and Functionalities 4 Functionality Goals 5 Threat Models and Security Requirements 6 High-Level Workflow 7 Example Protocol Instantiations 7.1 Distributed Data Discovery (MedCo) Setup Initialization ETL Process Query Generation Query Re-encryption Local Query Processing Result Obfuscation Result Shuffling Proxy Re-encryption of the Result Decryption 7.2 Centralized Data Analysis (Private Evaluation of Random Forests) 7.3 Distributed Data Analysis (Statistical Computation and Training of Machine Learning Models) 8 Concluding Remarks References Secure and Confidential Rule Matching for Network Traffic Analysis 1 Introduction 1.1 Motivation and Business Problem 2 Threat Model 3 Protocol 3.1 Client 3.2 Solution Provider 3.3 Rule Sets Examples of Rules 3.4 Prerequisites of the Protocol 3.5 Protocol Steps 4 Performance, Usability, and Scalability 4.1 Security Agencies 4.2 Fraud Detection References Trusted Monitoring Service (TMS) 1 Privacy-Preserving Health Monitoring 2 Business Motivation 3 Protocol (Workflow) 4 Performance, Usability, Scalability 5 Applications of Trusted Monitoring Systems References Private Set Intersection and Compute 1 Motivation 1.1 Privacy Compliance 1.2 Co-marketing as a Use Case 2 Application Functionality 2.1 Database Statistics on PSI Selected Entries 3 Protocol 3.1 Workflow 3.2 First Protocol: N Parties with One Central Compute Node 4 Examples 4.1 IXUP 4.2 Private Join and Compute 5 Performance, Usability, and Scalability References Part IV Applications of Homomorphic Encryption Private Outsourced Translation for Medical Data 1 Introduction 2 Machine Translation 3 Design 3.1 Challenges 4 Implementation and Evaluation 4.1 Encoding 4.2 Optimizations 4.3 Results 5 Discussion References HappyKidz: Privacy Preserving Phone Usage Tracking 1 Introduction 1.1 Privacy Model 2 Proof of Concept Implementation 2.1 Data Selection and Features 2.2 Learning Model 2.3 Microsoft SEAL Implementation 3 Soundness and Future Work 3.1 Future Work 4 Conclusion References i-SEAL2: Identifying Spam EmAiL with SEAL 1 Introduction 2 Private Classification 3 Private Training 4 Conclusion PRIORIS: Enabling Secure Detection of Suicidal Ideation from Speech Using Homomorphic Encryption 1 Introduction 2 Suicide Ideation Detection 2.1 Dataset 2.2 Application 3 Use Cases 3.1 Use-Case 1: Secure Detection and Response 3.2 Use-Case 2: Secure Clinical Assessment Assistance 3.3 Use-Case 3: Secure Treatment Evaluation 4 Network Training 5 Homomorphic Network Evaluation 6 Extensions and Future Work 7 Conclusion References Gimme That Model!: A Trusted ML Model Trading Protocol 1 Introduction 2 Non-cryptographic Approaches and Their Drawbacks 3 Our HE-Based Cryptographic Solution 3.1 The Protocol 3.2 Efficiency of the Protocol 3.3 Towards the Perfect Model Protection 3.4 Compatible ML Models 4 Discussions 4.1 Plausibility of Trading ML Models 4.2 Alternative Cryptographic Solutions 4.3 Dual Scenario: Trading Datasets References HEalth: Privately Computing on Shared Healthcare Data 1 Introduction and Motivation 2 Our Scenario 3 A Discussion of the Underlying Cryptography 4 The Initial Goal: Fairness 5 Discussion References Private Movie Recommendations for Children 1 Introduction 1.1 Background 2 Proposed Implementation 2.1 HE Technical Details 3 Discussion References Privacy-Preserving Prescription Drug Management Using Fully Homomorphic Encryption 1 Introduction 2 Our Model 3 Fully Homomorphic Encryption 3.1 Our Choice of FHE Scheme 3.2 Updating the Encrypted Records 3.3 Parameters 4 The Machine Learning Model 4.1 Training the Model 4.2 A Remark on Using ML 5 Authentication 5.1 The Shared Secret Key 5.2 Prevent Patient Tampering References Correction to: Introduction to Homomorphic Encryption and Schemes