/mphil_thesis_codes

Primary LanguageC++

Kernel Density Estimation on Homomorphically Encrypted Data

This repo contains the codes for my MPhil thesis.

Structure

  • folder main contains the functions used to create the figures
  • folder nonparametrics contains lower-level functions
  • once main.jl is run, folder figures will contain all figures
  • folders HEAAN-1.0 and ntl-11.4.3 might be required if HAAN.out needs to be recompiled (see below)

How to Run?

Make sure that julia is installed with the following packages:

  • Plots.jl
  • Distributions.jl
  • SpecialFunctions.jl
  • QuadGK.jl
  • JLD.jl
  • LaTeXStrings.jl
  • DelimitedFiles.jl

Once they are installed, clone this repo. If git is installed, this can be done from the git bash or terminal by the command git clone https://github.com/kmmate/mphil_thesis_codes.git.

Then from git bash or terminal cd mphil_thesis_codes, and then run julia ./main/src/main.jl to reproduce the thesis figures.

Notes

  • The executable main/src/HEAAN.out assumes that the machine has 8 cores. If this is not the case, HEAAN.cpp must be recompiled by g++. See main/src/HEAAN.cpp for more details.

  • HEAAN.out can be used in itself to compute HE-KDE on encrypted data. For this, replace the numbers in main/src/temp/x_data.txt with the non-encrypted i.i.d. sample and the numbers in main/src/temp/x_query.txt with the desired query points. Then run ./main/src/HEAAN.out [args], replacing [args] with the arguments described in main/src/HEAAN.cpp. The results will be written to main/src/temp/result.txt.