Summary of Backdoor Attack on Vertical Federated Graph Neural Network Learning, by Jirui Yang et al.
Backdoor Attack on Vertical Federated Graph Neural Network Learning
by Jirui Yang, Peng Chen, Zhihui Lu, Ruijun Deng, Qiang Duan, Jianping Zeng
First submitted to arxiv on: 15 Oct 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Artificial Intelligence (cs.AI); Cryptography and Security (cs.CR)
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 A novel Federated Graph Neural Network (FedGNN) is proposed to enable privacy-preserving training on distributed graph data, integrating federated learning with graph neural networks. The Vertical Federated Graph Neural Network (VFGNN), a key branch of FedGNN, handles scenarios where data features and labels are distributed among participants. However, VFGNN still faces the risk of backdoor attacks, even when labels are inaccessible. To address this, a new backdoor attack method called BVG is proposed, which leverages multi-hop triggers and backdoor retention to achieve nearly 100% attack success rates across three datasets and three GNN models with minimal impact on main task accuracy. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Federated Graph Neural Networks (FedGNN) are a way to train machines to work with big amounts of data without sharing the data itself. This is important for keeping information private. The paper shows that even with this system, there’s still a risk of someone putting fake information into the data. They call this “backdoor attacks.” To show how serious this problem is, they created a new way to do these backdoor attacks called BVG. They tested it on three different types of data and found that it worked almost every time. This means we need to find better ways to stop these kinds of attacks. |
Keywords
» Artificial intelligence » Federated learning » Gnn » Graph neural network
