On Strengthening and Defending Graph Reconstruction Attack with Markov Chain Approximation

Authors: Zhanke Zhou, Chenyu Zhou, Xuan Li, Jiangchao Yao, Quanming Yao, Bo Han

ICML 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Empirically, we achieve state-of-the-art results on six datasets and three common GNNs.
Researcher Affiliation Collaboration 1Department of Computer Science, Hong Kong Baptist University 2Cooperative Medianet Innovation Center, Shanghai Jiao Tong University 3Shanghai AI Laboratory 4Department of Electronic Engineering, Tsinghua Unversity.
Pseudocode Yes Algorithm 1 Markov Chain-based Graph Reconstruction Attack. Algorithm 2 Markov Chain-based Defensive Training Against Graph Reconstruction Attack.
Open Source Code Yes The code is publicly available at: https: //github.com/tmlr-group/MC-GRA.
Open Datasets Yes Six common datasets are utilized in experiments, which are collected from four diverse domains: (1) Cora and Citeseer (Sen et al., 2008) are citation networks... (2) Polblogs (Adamic & Glance, 2005) is a social network... (3) USA and Brazil (Ribeiro et al., 2017) are air-traffic networks... (4) AIDS (Riesen et al., 2008) is a chemical network...
Dataset Splits Yes The dataset split setting of train/validate/test sets is consistent with other datasets used in this work.
Hardware Specification Yes The implementation software is Pytorch (Paszke et al., 2017) while the hardware is an NVIDIA RTX 3090 GPU.
Software Dependencies No The implementation software is Pytorch (Paszke et al., 2017)... It mentions PyTorch but does not specify its version number.
Experiment Setup Yes The optimal hyperparameters are obtained by random search or grid search (La Valle et al., 2004).