Reproduce “Attack Graph Convolutional Networks by Adding Fake Nodes” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Xiaoyun Wang, Minhao Cheng, Joe Eaton, Cho-Jui Hsieh, Felix Wu
The authors introduce attacks that add fake nodes and links to disrupt GCN classification without changing existing nodes’ features or connections. Greedy selects binary links and features, while Greedy-GAN adds a discriminator to make generated features harder to distinguish. On Citeseer with symmetric normalization, reported accuracy falls from 0.73 to 0.03, demonstrating vulnerability on tested citation graphs, not deployed networks.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.