Summary of Defense-as-a-service: Black-box Shielding Against Backdoored Graph Models, by Xiao Yang et al.
Defense-as-a-Service: Black-box Shielding against Backdoored Graph Models
by Xiao Yang, Kai Zhou, Yuni Lai, Gaolei Li
First submitted to arxiv on: 7 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 GraphProt is a novel defense mechanism designed to protect GNN-based graph classifiers from backdoor attacks in large graph learning models. The approach allows resource-constrained business owners to rely on third-party models without compromising privacy. GraphProt is model-agnostic, only requiring the input graph as input. It comprises two components: clustering-based trigger elimination and robust subgraph ensemble. Experimental results across three backdoor attacks and six benchmark datasets demonstrate that GraphProt significantly reduces the backdoor attack success rate while preserving the model accuracy on regular graph classification tasks. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary GraphProt is a new way to keep GNN-based models safe from hacking. These models are used in big data learning, but sometimes bad guys can trick them into making wrong predictions. The current ways to stop this are limited and hard to use. GraphProt lets business owners use third-party models without worrying about privacy. It only needs the input graph and is not specific to any model. Two parts make it work: removing trigger graphs and creating a strong model using these removed graphs. |
Keywords
» Artificial intelligence » Classification » Clustering » Gnn
