Summary of Minimum Topology Attacks For Graph Neural Networks, by Mengmei Zhang et al.
Minimum Topology Attacks for Graph Neural Networks
by Mengmei Zhang, Xiao Wang, Chuan Shi, Lingjuan Lyu, Tianchi Yang, Junping Du
First submitted to arxiv on: 5 Mar 2024
Categories
- Main: Artificial Intelligence (cs.AI)
- Secondary: None
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary The proposed paper introduces a novel type of Graph Neural Network (GNN) attack, named minimum-budget topology attack, designed to adaptively find the minimum perturbation sufficient for successful misclassification of each node. This approach aims to overcome the limitations of traditional fixed-budget attacks, which can result in either no successful perturbations or redundant ones. The authors develop an attack model, MiBTack, based on a dynamic projected gradient descent algorithm, which effectively solves non-convex constraint optimization on discrete topology. Experimental results demonstrate the efficacy of MiBTack on three GNNs and four real-world datasets, showcasing its ability to successfully misclassify all target nodes with minimal perturbation edges. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary The paper proposes a new way to attack Graph Neural Networks (GNNs) called the minimum-budget topology attack. This method tries to find the smallest change needed to make each node in the graph do something different. The authors developed an algorithm, MiBTack, that can do this by adjusting some nodes and edges in the graph. They tested it on four real-world datasets and three types of GNNs and showed that it can successfully trick all the nodes into doing what they want. |
Keywords
» Artificial intelligence » Gnn » Gradient descent » Graph neural network » Optimization
