jeremykohn/iced-covid

COVID-19 diagnosis using encrypted medical images to preserve patient privacy. Links to software, research, and other resources.

Image Classification on Encrypted Data for Clinical Open-source Viral Infection Diagnosis (ICED-COVID)

The initial task is to compile and organize existing resources.

The main goal will be to create a "proof-of-concept" application that will train a machine learning classifier on encrypted images, run the classifier on additional encrypted images, and return the classifier's encrypted predictions.

Existing projects for encrypted image classification

Encrypted image classification

• minkcoregame/privacy-CNN

  • Project description: This repository provides python code for privacy preserving image classification based on fully homomorphic encryption (FHE).

• shreyagarge/EncryptedImagePredictorNetwork

  • Project description: Implementation of a Neural Network that predicts encrypted handwritten digits from the MNIST dataset. Encryption and Network Implemented using Homomorphic encryption library SEAL. Model obtained by training on unencrypted images using TensorFlow.
  • License: GNU General Public License (GPL) v3.0

• fentec-project/neural-network-on-encrypted-data

  • Project description: Demonstrating neural network model applied on encrypted data by using functional encryption.
  • License: Apache License 2.0

Homomorphic encryption of images

• Homomorphic-Image-Encryption

  • Project description: Secure Outsourcing of Image Editing Based on Homomorphic Encryption
  • License: GNU General Public License (GPL) v3.0

• adhilaq/Homomorphic-Image-Encryption

  • Project description: Performance and Security Analysis on following Homomorphic Encryption Algorithms

• Computer-Security-homomorphic-image-encryption

  • Project description: Implemented 3 homomorphic image encryption algorithms- RSA, Pallier & Gentry(Full) in Python and verified their authenticity by doing various image processing operations on the encrypted image.

• Image-Encryption-based-on-Paillier-Cryptosystem

  • Project description: Image Encryption using Homomorphic Encryption Scheme

Homomorphic encryption libraries

Library	Language	Description	License
concrete	Rust	Rust FHE library that implements Zama's variant of TFHE.	GNU Affero General Public License (AGPL) v3.0 (concrete/concrete/LICENSE)
cuFHE	C++	CUDA-accelerated Fully Homomorphic Encryption Library.	MIT License
cuHE	C++	GPU-accelerated HE library for NVIDIA CUDA-Enabled GPUs.	MIT License
Cupcake	Rust	Facebook's Rust library for the (additive version of the) Fan-Vercauteren scheme.	MIT License
cuYASHE	C++	Based on leveled fully HE scheme YASHE for GPGPUs.	GNU General Public License (GPL) v3.0
FHEW	C++	A Fully HE library based on FHEW: Bootstrapping Homomorphic Encryption in less than a second.	GNU General Public License (GPL) v2.0, or (at your option) any later version
FV-NFLlib	C++	A header-only library implementing the Fan-Vercauteren scheme.	GNU General Public License (GPL) v3.0
HEAAN	C++	Scheme with native support for fixed point approximate arithmetic.	Creative Commons Attribution-NonCommercial 3.0 Unported (CC BY-NC 3.0)
HEAAN-Python	Python	Python binding for the HEANN library.	MIT License
HElib	C++	BGV scheme with bootstrapping and the Approximate Number CKKS scheme.	Apache License 2.0
HEMat	C++	C++ implementation of matrix computation (addition, multiplication, and transposition) using HEANN.	MIT License
krypto	C++	C++ implementation of multivariate quadratic FHE.
Λ ○ λ	Haskell	"Lol" Haskell library for ring-based lattice cryptography that supports FHE.	GNU General Public License (GPL) v3.0 (Lol/lol/LICENSE)
lattigo	Go	Go library for lattice-based crypto that implements various schemes.	Apache License 2.0
libScarab	C	C library implementing a FHE scheme using large integers.	Educational or academic use.
libshe	C++	Symmetric somewhat HE library based on DGHV scheme.	GNU General Public License (GPL) v3.0
Microsoft SEAL	C++	C++ FHE library implementing BFV and CKKS schemes.	MIT License
NFLlib	C++	NTT-based Fast Lattice library specialized on power-of-two polynomials.	MIT License
node-seal	JavaScript, WebAssembly	JavaScript/WebAssembly port of Microsoft SEAL.	MIT License
NuFHE	Python	GPU-accelerated HE library, faster than cuFHE, that implements the tfhe algorithms.	GNU General Public License (GPL) v3.0
PALISADE	C++, with Python wrapper available	Efficient implementations of lattice cryptography building blocks and leading homomorphic encryption schemes.	BSD 2-Clause
petlib	Python	Python library that implements a number of Privacy Enhancing Technologies.	BSD 2-Clause
Pyfhel	Python	A Python wrapper for SEAL, HElib, and PALISADE.	GNU General Public License (GPL) v3.0
PySyft	Python	Python library for secure and private Deep Learning.	Apache License 2.0 (PySyft/packages/syft/LICENSE)
python-paillier	Python	Partially HE based on Paillier scheme.	GNU General Public License (GPL) v3.0
SEAL-python	Python	Python binding for the Microsoft SEAL library.	MIT License
SparkFHE	Apache Spark	Apache Spark with an add-on for FHE computations. See 📄.	Apache License 2.0
TenSEAL	Python	Library for HE operations on tensors, built on Microsoft SEAL, with a Python API.	Apache License 2.0
tfhe	C++	Faster fully HE: Bootstrapping in less than 0.1 seconds.	Apache License 2.0

Research

Overview

• Computing Arbitrary Functions of Encrypted Data. Craig Gentry. 2010. (PDF)

Surveys, reviews, and summaries of research

• Confidential Machine Learning on Untrusted Platforms: A Survey. Sagar Sharma, Keke Chen. 2020-12-15.

• A Systematic Comparison of Encrypted Machine Learning Solutions for Image Classification. Veneta Haralampieva, Daniel Rueckert, Jonathan Passerat-Palmbach. 2020-11-10.

Additional research papers and studies

• Efficient CNN Building Blocks for Encrypted Data. Nayna Jain, Karthik Nandakumar, Nalini Ratha, Sharath Pankanti, Uttam Kumar. 2021-01-30.

• On the Security of Homomorphic Encryption on Approximate Numbers. Baiyu Li, Daniele Micciancio. Received 2020-12-07, last revised 2021-03-07.

• CaRENets: Compact and Resource-Efficient CNN for Homomorphic Inference on Encrypted Medical Images. Jin Chao, Ahmad Al Badawi, Balagopal Unnikrishnan, Jie Lin, Chan Fook Mun, James M. Brown, J. Peter Campbell, Michael Chiang, Jayashree Kalpathy-Cramer, Vijay Ramaseshan Chandrasekhar, Pavitra Krishnaswamy, Khin Mi Mi Aung. 2019-01-29.

• CryptoNets: Applying Neural Networks to Encrypted Data with High Throughput and Accuracy. Nathan Dowlin, Ran Gilad-Bachrach, Kim Laine, Kristin Lauter, Michael Naehrig, John Wernsing. Microsoft Research, Redmond, USA; Princeton University, New-Jersey, USA. 2016-02-24.

Other resources

• Homomorphic Encryption Standardization

  • Description: an open consortium of industry, government and academia to standardize homomorphic encryption.

• Awesome Homomorphic Encryption

  • Desctiption: A curated list of amazing Homomorphic Encryption libraries, software and resources.

• CrypTen

  • Description: A research tool for secure machine learning in PyTorch. CrypTen currently implements a cryptographic method called secure multiparty computation (MPC), and we expect to add support for homomorphic encryption and secure enclaves in futue releases.
  • Website: https://crypten.ai/

• ZDNet: AI runs smack up against a big data problem in COVID-19 diagnosis (By Tiernan Ray, April 4, 2020)

  • Researchers around the world have quickly pulled together combinations of neural networks that show real promise in diagnosing COVID-19 from chest X-rays and CT scans. But a lack of data is hampering the ability of many efforts to move forward. Some kind of global data sharing may be the answer.