A small repository to contain all the resources of Multi-Party Computation and Differential Privacy used by me to develop background for IBM Research collab project.
[Update] Note: This will not contain any of the research work in general; just all the literature-reading.
-
Secure Multiparty Computation (MPC) by Yehuda Lindell (cryptographic protocols / survey): Paper | Video
-
MP-SPDZ: A Versatile Framework for Multi-Party Computation by Marcel Keller : Paper
-
SCALE-MAMBA Documentation & Guide : link
-
Morten Dahl's blogs on MPC and SPDZ Protocol : link
-
Bristol Cryptography Blog : link
-
A Pragmatic Introduction to Secure Multi-Party Computation: PDF
-
IBM's Differential Privacy library
diffprivlib
: installation guide | example notebooks | detailed documentation -
Google's Differential Privacy library
Privacy on Beam
: documentation | codelab | Github -
OpenMined's Python wrapper for Privacy on Beam
PyDP
: Github | Slack | excellent examples | resources & tutorials -
Facebook's high-speed library for training PyTorch models with differential privacy
Opacus
: Tutorial
-
k-anonymity : Wikipedia
-
OpenDP's
SmartNoise SDK
: GitHub -
Google's Differential Privacy library Algo's documentation (C++) : Docu
-
TED's non-technical writing on Differential Privacy : Link
- Frankmcsherry's Blog : Link
- Uber's Differential Privacy Framework (based on Elastic Sensitivity)
CHORUS
: First Blog | Updated Blog | Video presentation | Elastic Sensitivity paper | GitHub
-
Hierarchial Time Series: Tutorials
-
Vector autoregression : Wikipedia | Tutorials | PennState Docu
-
ARIMA model (Autoregressive Integrated moving averages : Wikipedia | Tutorials
-
Working mechanisms of different DP libaries : Privacy on Beam | OpenDP
-
Distributed DP mechanisms into MPC protocols: Paper 1 | Paper 2 | Paper 3
-
Laplace mechanism for adding noise in ε-differential privacy: Paper 1 | Slides from Roger Grosse | Lecture of Gautam Kamath | Truncated Laplacian mechanism used by IBM
-
Geometric mechanism used in
IBM diffprivlib
: Paper -
DLPA (Distributed Laplace Perturbation Algorithm) for optimally adding noise to attain Differential Privacy in a distributed manner when N users are participating in a computation: Paper (Rastogi and Nath)
-
The parameters used in
IBM diffprivlib
for calculating differentially private sum/mean: Reference Pg 41,44
-
How to sample noise from Gaussian Distribution : 3 standard .py libs | numpy.random Docu
-
MP-SPDZ for Secure MPC : Github | tutorial.mpc | mp-spdz documentation
-
Different Functions(sum, mean, count, var, sd, quantiles) : IBM diffprivlib utils.py | IBM diffprivlib quantiles.py
-
Meaning of Epsilon(ε) and Delta(δ) in (ε, δ)-differential privacy : Ref