MIA GNN Project Starter

If you meet the version mismatch error for Lasagne library, please use following command to upgrade Lasagne library.

pip install --upgrade https://github.com/Lasagne/Lasagne/archive/master.zip

MIA Attack Process

Step1: Training Target Model using Target Dataset
Step2: Training Shadow Model using Shadow Dataset
Step3: Training Attack Model using Posteriors retrieved from Shadow Model

Here, Target Dataset and Shadow Dataset are disjoint.

Training Target and Shadow Model by GCN model

TUs: DD, PROTEINS_full, ENZYMES

# 10: run 10 times ;100:start from 100 epochs; DD : dataset DD
sh run_TUs_target_shadow_training.sh 10 100 DD

SPs: CIFAR10, MNIST

# 10: run 10 times ;100:start from 100 epochs; DD : dataset DD
sh run_SPs_target_shadow_training.sh 10 100 DD

Membership Inference Attack

# For transfer based attack, run 15 times
sh run_transfer_attach.sh 15

Acknowledge

This project references from benchmarking-gnn and DeeperGCN

If it has any issues, please let me know.

Membership Inference Attack against Graph Neural Networks

Related tags

Overview

MIA GNN Project Starter

MIA Attack Process

Training Target and Shadow Model by GCN model

TUs: DD, PROTEINS_full, ENZYMES

SPs: CIFAR10, MNIST

Membership Inference Attack

Acknowledge

Owner

SlideGraph+: Whole Slide Image Level Graphs to Predict HER2 Status in Breast Cancer

Reading Group @mila-iqia on Computational Optimal Transport for Machine Learning Applications

This code finds bounding box of a single human mouth.

Code for the paper: Adversarial Training Against Location-Optimized Adversarial Patches. ECCV-W 2020.

Running Google MoveNet Multipose Tracking models on OpenVINO.

A deep learning library that makes face recognition efficient and effective

Uses Open AI Gym environment to create autonomous cryptocurrency bot to trade cryptocurrencies.

Computationally Efficient Optimization of Plackett-Luce Ranking Models for Relevance and Fairness

We propose a new method for effective shadow removal by regarding it as an exposure fusion problem.

Course on computational design, non-linear optimization, and dynamics of soft systems at UIUC.

Notebooks, slides and dataset of the CorrelAid Machine Learning Winter School

PaddleRobotics is an open-source algorithm library for robots based on Paddle, including open-source parts such as human-robot interaction, complex motion control, environment perception, SLAM positioning, and navigation.

CvT-ASSD: Convolutional vision-Transformerbased Attentive Single Shot MultiBox Detector (ICTAI 2021 CCF-C 会议)The 33rd IEEE International Conference on Tools with Artificial Intelligence

RM Operation can equivalently convert ResNet to VGG, which is better for pruning; and can help RepVGG perform better when the depth is large.

PIXIE: Collaborative Regression of Expressive Bodies

The VarCNN is an Convolution Neural Network based approach to automate Video Assistant Referee in football.

Pytorch domain adaptation package

A web porting for NVlabs' StyleGAN2, to facilitate exploring all kinds characteristic of StyleGAN networks

ROSITA: Enhancing Vision-and-Language Semantic Alignments via Cross- and Intra-modal Knowledge Integration

PyTorch implementation of the end-to-end coreference resolution model with different higher-order inference methods.