A curated list of resources for privacy-preserving machine learning.
See also:
- awesome-he - for homomorphic encryption
- awesome-mpc - for secure multi-party computation
- awesome-differential-privacy - for differential privacy
which also contain links to some of the (more general purpose) tools often used in with PPML.
- PPML News and updates on Twitter
- IACR ePrint archive and updates on Twitter
- Cryptography and Security on arXiv.org
- Machine Learning on arXiv.org
- HE Transformer - homomorphic encryption backend for nGraph
- TensorFlow Privacy - differential privacy in TensorFlow
- tf-encrypted - secure multi-party computation in TensorFlow
- PySyft - secure multi-party computation and differential privacy in PyTorch
- Hacking Deep Learning
- Privacy-Preserving Machine Learning, NeurIPS'18
- Private Multi-Party Machine Learning, NIPS'16
A great summary is provided in MRSV'17 and the archives of PPML News contain more papers in chronological order.
Selection:
- Faster CryptoNets: Leveraging Sparsity for Real-World Encrypted Inference, CBLYHF'18
- nGraph-HE: A Graph Compiler for Deep Learning on Homomorphically Encrypted Data, BLW'18
- CHET: Compiler and Runtime for Homomorphic Evaluation of Tensor Programs, DSCLLMMM'18
- Slalom: Fast, Verifiable and Private Execution of Neural Networks in Trusted Hardware, TB'18
- SecureNN: Efficient and Private Neural Network Traning, WGC'18
- ABY3: A Mixed Protocol Framework for Machine Learning, MR'18
- Chiron: Privacy-preserving Machine Learning as a Service, HSSSW'18
- Scalable Private Learning with PATE, PSMRTE'18
- EPIC: Efficient Private Image Classification, MRSV'17
- Gazelle: A Low Latency Framework for Secure Neural Network Inference, JVC'18
- Chameleon: A Hybrid Secure Computation Framework for Machine Learning Applications, RWTSSK'17
- DeepSecure: Scalable Provably-Secure Deep Learning, RRK'17
- Oblivious Neural Network Predictions via MiniONN transformations, LJLA'17
- SecureML: A System for Scalable Privacy-Preserving Machine Learning, MZ'17
- CryptoNets: Applying Neural Networks to Encrypted Data with High Throughput and Accuracy, DGLLNW'16
- Andrew Trask and OpenMined community
- Adrià Gascón