Reliable Graph Neural Networks via Robust Aggregation
arXiv:2010.15651
Abstract
Perturbations targeting the graph structure have proven to be extremely effective in reducing the performance of Graph Neural Networks (GNNs), and traditional defenses such as adversarial training do not seem to be able to improve robustness. This work is motivated by the observation that adversarially injected edges effectively can be viewed as additional samples to a node's neighborhood aggregation function, which results in distorted aggregations accumulating over the layers. Conventional GNN aggregation functions, such as a sum or mean, can be distorted arbitrarily by a single outlier. We propose a robust aggregation function motivated by the field of robust statistics. Our approach exhibits the largest possible breakdown point of 0.5, which means that the bias of the aggregation is bounded as long as the fraction of adversarial edges of a node is less than 50\%. Our novel aggregation function, Soft Medoid, is a fully differentiable generalization of the Medoid and therefore lends itself well for end-to-end deep learning. Equipping a GNN with our aggregation improves the robustness with respect to structure perturbations on Cora ML by a factor of 3 (and 5.5 on Citeseer) and by a factor of 8 for low-degree nodes.
23 pages, 9 figures, 6 Tables, Neural Information Processing Systems, NeurIPS, 2020
References in corpus (10)
- Inductive Representation Learning on Large Graphs
- Graph Neural Networks: A Review of Methods and Applications
- Adversarial Attacks on Node Embeddings via Graph Poisoning
- Adversarial Attack on Graph Structured Data
- Adversarial Examples on Graph Data: Deep Insights into Attack and Defense
- Recent Advances in Algorithmic High-Dimensional Robust Statistics
- Topology Attack and Defense for Graph Neural Networks: An Optimization Perspective
- Minimizing Polynomial Functions
- Geometric Median in Nearly Linear Time
- Topological Effects on Attacks Against Vertex Classification