Summary of Hgattack: Transferable Heterogeneous Graph Adversarial Attack, by He Zhao et al.
HGAttack: Transferable Heterogeneous Graph Adversarial Attack
by He Zhao, Zhiwei Zeng, Yongwei Wang, Deheng Ye, Chunyan Miao
First submitted to arxiv on: 18 Jan 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Cryptography and Security (cs.CR); Information Retrieval (cs.IR)
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 This paper introduces a new method for generating adversarial attacks on Heterogeneous Graph Neural Networks (HGNNs). Existing methods, designed for homogeneous graphs, fail to address the structural and semantic complexity of HGNNs. The proposed approach, called HGAttack, uses a novel surrogate model that extracts meta-path induced subgraphs and applies GNNs to learn node embeddings with distinct semantics. This improves the transferability of generated attacks on the target HGNN and reduces memory costs. For perturbation generation, HGAttack leverages subgraph gradient information to identify vulnerable edges across various relations within a constrained budget. The paper validates HGAttack’s efficacy through comprehensive experiments on three datasets, demonstrating significant performance degradation for target HGNN models. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary HGAttack is a new method that helps protect Heterogeneous Graph Neural Networks from bad guys trying to mess with them. These attacks are like trying to trick a computer into doing the wrong thing. The old ways of attacking didn’t work well for these special kinds of networks, so researchers created HGAttack to help fix that. It works by making a copy of the network that is similar but not exactly the same, and then tries to find the weak spots in the original network. This helps make the original network stronger against attacks. The researchers tested HGAttack on three different datasets and it did a great job of making the networks harder to trick. |
Keywords
* Artificial intelligence * Semantics * Transferability
