Giter Site home page Giter Site logo

lycogno / mia-icpr Goto Github PK

View Code? Open in Web Editor NEW
1.0 1.0 0.0 438 KB

This repository accompanies the paper "SynthShield: Leveraging Synthetic Distributions to Enhance Privacy Against Membership Inference" currently under review at the International Conference on Pattern Recognition (ICPR). It contains the main code used in applying and analysing the SynthShield technique analysed in the paper.

License: Other

Python 100.00%
generative-ai machine-learning membership-inference-attack privacy

mia-icpr's Introduction

SynthShield: Leveraging Synthetic Distributions to Enhance Privacy Against Membership Inference

This repository accompanies the paper "SynthShield: Leveraging Synthetic Distributions to Enhance Privacy Against Membership Inference" currently under review at the International Conference on Pattern Recognition (ICPR). It contains the main code used in applying and analysing the SynthShield technique analysed in the paper. Our code is based on Membership Inference Attacks and Defenses in Neural Network Pruning. Testing was done on Ubuntu 20.04.6 LTS with Python 3.10.9.

Our technique

Background

The growing application of machine learning on Personally Identifiable Information (PII) raises various privacy concerns. Techniques known as Membership Inference Attacks (MIAs) can be used to extract information from a trained model regarding its dataset. Previous research has shown that common pruning techniques increase model vulnerability to such attacks. Existing defenses to solve this include regularisation-based approaches, transfer learning, and perturbation-based approaches, among others. In this paper, we provide the first in-depth analysis of the effect of fine-tuning on a synthetic dataset generated to mimic the original distribution on MIA attack accuracy. We further show its effectiveness in defending against various black-box SOTA membership inference attacks.

Installation

Clone the repo

git clone https://github.com/lycogno/mia-icpr
cd mia-icpr

Install packages (Python)

pip install -r requirements.txt

Make directories for outputs

mkdir saved_models_afterprune
mkdir shadow_models_afterprune

NOTE: The repo expects to find CIFAR10 data in cifardata/ and generated data in data/.

Commands

Generate victim models with our proposed defense

python prune.py --classifier mobilenet_v2

Generate shadow models

python prune_only_forshadow.py --classifier mobilenet_v2

Run the ablation

python mia_ours_ablation.py 0 config/cifar10_resnet18.json --attacks samia,nn,nn_top3 --classifier mobilenet_v2

mia-icpr's People

Contributors

lycogno avatar

Stargazers

 avatar

Watchers

 avatar

Recommend Projects

  • React photo React

    A declarative, efficient, and flexible JavaScript library for building user interfaces.

  • Vue.js photo Vue.js

    ๐Ÿ–– Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.

  • Typescript photo Typescript

    TypeScript is a superset of JavaScript that compiles to clean JavaScript output.

  • TensorFlow photo TensorFlow

    An Open Source Machine Learning Framework for Everyone

  • Django photo Django

    The Web framework for perfectionists with deadlines.

  • D3 photo D3

    Bring data to life with SVG, Canvas and HTML. ๐Ÿ“Š๐Ÿ“ˆ๐ŸŽ‰

Recommend Topics

  • javascript

    JavaScript (JS) is a lightweight interpreted programming language with first-class functions.

  • web

    Some thing interesting about web. New door for the world.

  • server

    A server is a program made to process requests and deliver data to clients.

  • Machine learning

    Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.

  • Game

    Some thing interesting about game, make everyone happy.

Recommend Org

  • Facebook photo Facebook

    We are working to build community through open source technology. NB: members must have two-factor auth.

  • Microsoft photo Microsoft

    Open source projects and samples from Microsoft.

  • Google photo Google

    Google โค๏ธ Open Source for everyone.

  • D3 photo D3

    Data-Driven Documents codes.