Adversarial Attacks on Node Embeddings via Graph Poisoning

By Aleksandar Bojchevski, Stephan Günnemann. 2018

350 citations recorded by the source corpus.

The goal of network representation learning is to learn low-dimensional node embeddings that capture the graph structure and are useful for solving downstream tasks. However, despite the proliferation of such methods, there is currently no study of their robustness to adversarial attacks. We provide the first adversarial vulnerability analysis on the widely used family of methods based on random walks. We derive efficient adversarial perturbations that poison the network structure and have a negative effect on both the quality of the embeddings and the downstream tasks. We further show that our attacks are transferable since they generalize to many models and are successful even when the attacker is restricted.

Subfields: Artificial Intelligence, Computer Networks and Communications

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. Neil Zhenqiang Gong
2. Binghui Wang
3. Marinka Zitnik
4. Lichao Sun
5. Philip S. Yu
6. Stephan Günnemann
7. Daniel Zügner
8. Yingtong Dou

In Computer Networks and Communications

1. Binghui Wang
2. Neil Zhenqiang Gong
3. Daniel Zügner
4. Marinka Zitnik
5. Wei Yang
6. Anki Chauhan
7. Mirazul Haque
8. Cong Liu

Citations in the corpus

• Attacking Graph-based Classification via Manipulating the Graph Structure (2019, background): “Several work [8, 16, 57, 77, 78] proposed attacks to graph neural network methods via inserting fake edges to the graph or deleting existing edges.”
• GNNGuard: Defending Graph Neural Networks against Adversarial Attacks (2020, background): “[30] derive adversarial perturbations that poison the graph structure.”
• Adversarial Attack and Defense on Graph Data: A Survey (2018, methodology): “to the model and make the model inaccurate when using the modified information [8], [28], [145].”
• Adversarial Attacks on Graph Neural Networks via Meta Learning (2019, background): “…classical approaches to node classification (London & Getoor, 2014; Chapelle et al., 2006), recently deep learning on graphs has gained much attention (Monti et al., 2017; Bojchevski & Günnemann, 2018a; Battaglia et al., 2018; Perozzi et al., 2014; Bojchevski et al., 2018; Klicpera et al., 2019).”
• ILFO: Adversarial Attack on Adaptive Neural Networks (2020, methodology): “With a similar approach, adversarial attacks have been extended to various fields like text and speech processing [3, 18], and graph models [31, 1].”
• Certifiable Robustness and Robust Training for Graph Convolutional Networks (2019, background): “X̂q,Q ( Û X ) = { X̃ X̃nj ∈ [0, 1] ∧ ∥X̃ − Û X ∥1 ≤ Q (8)”
• Adversarial Attacks and Defenses in Images, Graphs and Text: A Review (2019, neutral): “Their approach is based on meta learning[101], which is traditionally used for hyperparameter optimization, few-shot image recognition, and fast reinforcement learning.”
• Graph Structure Learning for Robust Graph Neural Networks (2020, neutral): “GNNs also suffer this problem [5, 10, 17, 23, 24, 32, 39, 40].”