Explainability-based Backdoor Attacks Against Graph Neural Networks

Abstract

Backdoor attacks represent a serious threat to neural network models. A backdoored model will misclassify the trigger-embedded inputs into an attacker-chosen target label while performing normally on other benign inputs. There are already numerous works on backdoor attacks on neural networks, but only a few works consider graph neural networks (GNNs). As such, there is no intensive research on explaining the impact of trigger injecting position on the performance of backdoor attacks on GNNs. To bridge this gap, we conduct an experimental investigation on the performance of backdoor attacks on GNNs. We apply two powerful GNN explainability approaches to select the optimal trigger injecting position to achieve two attacker objectives -- high attack success rate and low clean accuracy drop. Our empirical results on benchmark datasets and state-of-the-art neural network models demonstrate the proposed method's effectiveness in selecting trigger injecting position for backdoor attacks on GNNs. For instance, on the node classification task, the backdoor attack with trigger injecting position selected by GraphLIME reaches over $84\%$ attack success rate with less than $2.5\%$ accuracy drop

arXiv comment6 pages, 4 figures

References in corpus (5)

• Inductive Representation Learning on Large Graphs
• Axiomatic Attribution for Deep Networks
• TUDataset: A collection of benchmark datasets for learning with graphs
• Anti-Money Laundering in Bitcoin: Experimenting with Graph Convolutional Networks for Financial Forensics
• Backdoor Attacks to Graph Neural Networks

Cited by in corpus (4)

• Model Extraction Attacks on Graph Neural Networks: Taxonomy and Realization
• Adversarial Attacks on Graph Classification via Bayesian Optimisation
• Accumulative Poisoning Attacks on Real-time Data
• Oriole: Thwarting Privacy against Trustworthy Deep Learning Models