GNNGuard: Defending Graph Neural Networks against Adversarial Attacks

By Marinka Zitnik. Neural Information Processing Systems, 2020

397 citations recorded by the source corpus.

Deep learning methods for graphs achieve remarkable performance on many tasks. However, despite the proliferation of such methods and their success, recent findings indicate that small, unnoticeable perturbations of graph structure can catastrophically reduce performance of even the strongest and most popular Graph Neural Networks (GNNs). Here, we develop GNNGuard, a general defense approach against a variety of training-time attacks that perturb the discrete graph structure. GNNGuard can be straightforwardly incorporated into any GNN. Its core principle is to detect and quantify the relationship between the graph structure and node features, if one exists, and then exploit that relationship to mitigate negative effects of the attack. GNNGuard uses network theory of homophily to learn how best assign higher weights to edges connecting similar nodes while pruning edges between unrelated nodes. The revised edges then allow the underlying GNN to robustly propagate neural messages in the graph. GNNGuard introduces two novel components, the neighbor importance estimation, and the layer-wise graph memory, and we show empirically that both components are necessary for a successful defense. Across five GNNs, three defense methods, and four datasets, including a challenging human disease graph, experiments show that GNNGuard outperforms existing defense approaches by 15.3% on average. Remarkably, GNNGuard can effectively restore the state-of-the-art performance of GNNs in the face of various adversarial attacks, including targeted and non-targeted attacks.

Subfields: Artificial Intelligence, Materials Chemistry

View Paper: https://www.semanticscholar.org/paper/73366d75289c5e37481639fb54fdee28a664e2b3

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

1. Ajmal Saeed Mian
2. Lichao Sun
3. Philip S. Yu
4. Mubarak Shah
5. Naveed Akhtar
6. Navid Kardan
7. Bo Li
8. Yingtong Dou

In Materials Chemistry

1. Bo Li
2. Philip S. Yu

Citations in the corpus

• Adversarial Attack and Defense on Graph Data: A Survey (2018, background): “[138] introduced neighbor importance estimation and the layer-wise graph memory components which can be integrated with GNNs.”
• Graph Neural Networks for Graphs With Heterophily: A Survey (2022, neutral): “8), driving defense strategies: • GNNGUARD [121]: Detects feature-structure correlations to filter adversarial heterophilic edges, ensuring robust message propagation.”
• Sequential Recommendation with Graph Neural Networks (2021, background): “Common options for kernel-basedmethods include cosine distance [30, 41], Euclidean distance [38, 43] and Mahalanobis distance [18, 34].”
• Advances in adversarial attacks and defenses in computer vision: A survey (2021, background): “Defending Graph Neural Networks (GNNs) is studied in [434].”