Awesome Secure-Machine Learning

A curated list of source of secure machine learning. (Feel Free to send a pull request if you have new papers to add.)

Homomorphic Encryption

Survey papers introducing HE:

• A Survey on Homomorphic Encryption Schemes: Theory and Implementation - Introduing the development of HE and also several popular HE schemas.
• Survey on Homomorphic Encryptionand Address of New Trend - With some introduction of HE's application recently.

HE schemes:

• Paillier(SomeWhat HE) - Public-Key Cryptosystems Based on Composite Degree Residuosity Classes
• BGV - Fully Homomorphic Encryption without Bootstrapping
• BV - Fully Homomorphic Encryption from Ring-LWE and Security for Key Dependent Messages
• FV - Somewhat Practical Fully HomomorphicEncryption
• GSW - Homomorphic Encryption from Learning with Errors:Conceptually-Simpler, Asymptotically-Faster, Attribute-Based
• TFHE - TFHE: Fast Fully Homomorphic Encryptionover the Torus
• HEAAN(CKKS) - Homomorphic Encryptionfor Arithmetic of Approximate Numbers
• HEAAN with bootstrap -Bootstrapping for ApproximateHomomorphic Encryption

More advances in HE:

• Takeshita et. al. -Enabling Faster Operations for Deeper Circuitsin Full RNS Variants of FV-like SomewhatHomomorphic Encryption
• Chen et.al. -Efficient Homomorphic ConversionBetween (Ring) LWE Ciphertexts

HE for private deep neural netowrk inference:

• CryptoNets - CryptoNets: Applying Neural Networks to Encrypted Datawith High Throughput and Accuracy
• FHE-DiNN - Fast Homomorphic Evaluation ofDeep Discretized Neural Networks
• Faster CryptoNets - Faster CryptoNets: Leveraging Sparsity forReal-World Encrypted Inference
• Chimera - Chimera: a unified framework for B/FV, TFHE and HEAAN fully homomorphic encryption andpredictions for deep learning
• TAPAS - TAPAS: Tricks to Accelerate (encrypted) Prediction As a Service
• nGraph-HE - nGraph-HE: A Graph Compiler for Deep Learning onHomomorphically Encrypted Data
• CryptoDL - Low Latency Privacy Preserving Inference
• SHE -SHE: A Fast and Accurate Deep Neural Network forEncrypted Data
• ZAMA -New Challenges for Fully Homomorphic Encryption

HE for private deep neural netowrk training:

• FHESGD - Towards Deep Neural Network Training on Encrypted Data
• Glyph - Glyph: Fast and Accurately Training Deep Neural Networks on Encrypted Data

HE for other models:

• Hardy et. al.(Logistic Regression) - Private federated learning on vertically partitioned datavia entity resolution and additively homomorphicencryption
• Crawford et. al. (Logistic Regression) - Doing Real Work with FHE: The Case of Logistic Regression
• SecureBoost (XGBoost) - SecureBoost: A Lossless Federated Learning Framework
• Cheon et. al. (Cluster) - Towards a Practical Cluster Analysisover Encrypted Data
• Jaschke et. al. (Cluster) - Unsupervised Machine Learning on Encrypted Data
• Han et. al. (Cluster) - Efficient Logistic Regression on Large EncryptedData
• Li et. al. (Distributed Logistic Regression) - Faster Secure Data Mining via Distributed Homomorphic Encryption

Multiparty Secure Computation

• Sameer Narahari Wagh - New Directions in EfficientPrivacy­Preserving Machine Learning

Garbled Circuit:

• Yao - Protocols for Secure Computations
• FairplayMP - FairplayMP: a system for secure multi-party computation
• Kolesnikov et. al. - Improved Garbled Circuit: Free XOR Gates andApplications
• Mohassel et. al. - Fast and Secure Three-party Computation:The Garbled Circuit Approach
• Zahur et. al. - Two Halves Make a WholeReducing Data Transfer in Garbled Circuits using Half Gates

Function Secret Sharing:

• Boyle et. al. - Function Secret Sharing
• Boyle et. al. - Function Secret Sharing: Improvements and Extensions.
• Boyle et. al. - Secure Computation with Preprocessing viaFunction Secret Sharing

Oblivious Transfer:

• Asharov et. al. - More Efficient Oblivious Transfer andExtensions for Faster Secure Computation*
• Ishai et. al. - ExtendingObliviousTransfersE±ciently

MPC for Deep Learning:

• SecureML - SecureML: A System for Scalable Privacy-PreservingMachine Learning
• EzPC - EzPC: Programmable, Efficient, and ScalableSecure Two-Party Computation for Machine Learning
• GAZELLE - GAZELLE: A Low Latency Framework for SecureNeural Network Inference
• ABY3 - ABY3: A Mixed Protocol Framework for Machine Learning
• SecureNN - SecureNN: 3-Party Secure Computation forNeural Network Training
• XONN - XONN:XNOR-based Oblivious Deep Neural Network Inference
• Flash - FLASH: Fast and Robust Framework forPrivacy-preserving Machine Learning
• ASTRA - ASTRA: High Throughput 3PC over Rings with Application to Secure Prediction
• Trident - Trident: Efficient 4PC Framework for PrivacyPreserving Machine Learning
• BLAZE - BLAZE: Blazing Fast Privacy-Preserving MachineLearning
• DELPHI - DELPHI: A Cryptographic Inference Service for Neural Networks
• FALCON - FALCON: Honest-Majority Maliciously Secure Frameworkfor Private Deep Learning
• AriaNN - ARIANN: Low-Interaction Privacy-PreservingDeep Learning via Function Secret Sharing
• SWIFT - SWIFT: Super-fast and Robust Privacy-Preserving Machine Learning

MPC for other models:

• SANNS(KNN) - SANNS:Scaling Up Secure Approximatek-Nearest Neighbors Search