HyperNear: Unnoticeable Node Injection Attacks on Hypergraph Neural Networks
Authors: Tingyi Cai, Yunliang Jiang, Ming Li, Lu Bai, Changqin Huang, Yi Wang
ICML 2025 | Venue PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Extensive experiments on five real-world datasets demonstrate Hyper Near s effectiveness, generalization, and stealth, outperforming baseline methods. |
| Researcher Affiliation | Academia | 1Zhejiang Key Laboratory of Intelligent Education Technology and Application, Zhejiang Normal University, Jinhua, China 2China-Mozambique Belt and Road Joint Laboratory on Smart Agriculture, Zhejiang Normal University, Jinhua, China 3School of Computer Science and Technology, Zhejiang Normal University, Jinhua, China 4School of Information Engineering, Huzhou University, Huzhou, China 5Zhejiang Institute of Optoelectronics, Jinhua, China 6Beijing Normal University, Beijing, China. Correspondence to: Yunliang Jiang <EMAIL>. |
| Pseudocode | Yes | Algorithm 1 Hyper Near: Node Inj Ection Attack on Hype Rgraph |
| Open Source Code | Yes | Our code is available at https://github.com/ca1man-2022/ Hyper Near. |
| Open Datasets | Yes | We evaluate our models on five real-world hypergraph datasets for hypernode classfication tasks, including DBLP (Rossi & Ahmed, 2015), Pubmed, Citeseer and Cora (Sen et al., 2008). |
| Dataset Splits | No | The paper mentions using 'the same preprocessed hypergraphs as those provided in the official implementations of Hyper GCN (Yadati et al., 2019) and Uni GNN (Huang & Yang, 2021)', but it does not explicitly provide specific percentages, counts, or a detailed methodology for these dataset splits within the text. |
| Hardware Specification | Yes | All experiments are conducted on a device with AMD EPYC 7543 32-core processor and a NVIDIA RTX A6000 GPU with 48 GB of RAM. |
| Software Dependencies | No | The paper does not explicitly mention specific software dependencies with version numbers (e.g., Python, PyTorch, TensorFlow, CUDA versions). |
| Experiment Setup | Yes | The default hyperparameter settings in our method are as follows, the parameter random seed s is 4202, the choice of hyperedge k is 10%, the ratio of the injected nodes α is 5%, the effect of controlling the loss of homophily λ and τ is 0.1. |