Projective Ranking-based GNN Evasion Attacks

Abstract

Graph neural networks (GNNs) offer promising learning methods for graph-related tasks. However, GNNs are at risk of adversarial attacks. Two primary limitations of the current evasion attack methods are highlighted: (1) The current GradArgmax ignores the "long-term" benefit of the perturbation. It is faced with zero-gradient and invalid benefit estimates in certain situations. (2) In the reinforcement learning-based attack methods, the learned attack strategies might not be transferable when the attack budget changes. To this end, we first formulate the perturbation space and propose an evaluation framework and the projective ranking method. We aim to learn a powerful attack strategy then adapt it as little as possible to generate adversarial samples under dynamic budget settings. In our method, based on mutual information, we rank and assess the attack benefits of each perturbation for an effective attack strategy. By projecting the strategy, our method dramatically minimizes the cost of learning a new attack strategy when the attack budget changes. In the comparative assessment with GradArgmax and RL-S2V, the results show our method owns high attack performance and effective transferability. The visualization of our method also reveals various attack patterns in the generation of adversarial samples.

arXiv commentAccepted by IEEE Transactions on Knowledge and Data Engineering

References in corpus (17)

• Neural Message Passing for Quantum Chemistry
• Adversarial Attacks on Neural Networks for Graph Data
• Hierarchical Graph Representation Learning with Differentiable Pooling
• GNNExplainer: Generating Explanations for Graph Neural Networks
• Anomaly Detection on Attributed Networks via Contrastive Self-Supervised Learning
• TUDataset: A collection of benchmark datasets for learning with graphs
• Poisoning Attacks to Graph-Based Recommender Systems
• Parameterized Explainer for Graph Neural Network
• Multivariate Time Series Forecasting with Dynamic Graph Neural ODEs
• Hierarchical Graph Pooling with Structure Learning
• Graph Neural Networks for Graphs with Heterophily: A Survey
• Distance Encoding: Design Provably More Powerful Neural Networks for Graph Representation Learning
• A Survey of Adversarial Learning on Graphs
• Attack Graph Convolutional Networks by Adding Fake Nodes
• Adversarial Attack on Hierarchical Graph Pooling Neural Networks
• From Unsupervised to Few-shot Graph Anomaly Detection: A Multi-scale Contrastive Learning Approach
• Spatio-Temporal Joint Graph Convolutional Networks for Traffic Forecasting

Cited by in corpus (2)

• Learning Strong Graph Neural Networks with Weak Information
• Dynamic Graph Unlearning: A General and Efficient Post-Processing Method via Gradient Transformation