Robust Graph Contrastive Learning with Information Restoration

TL;DR

This paper introduces a robust graph contrastive learning framework that restores mutual information diminished by structural attacks, significantly improving resilience of graph embeddings without requiring label information.

Contribution

We propose a novel unsupervised method with a learnable sanitation view to enhance GCL robustness against structural attacks by restoring mutual information.

Findings

Our method outperforms baselines in robustness against structural attacks.

Restoring mutual information improves embedding quality under attack.

Unsupervised hyperparameter tuning is effective without label access.

Abstract

The graph contrastive learning (GCL) framework has gained remarkable achievements in graph representation learning. However, similar to graph neural networks (GNNs), GCL models are susceptible to graph structural attacks. As an unsupervised method, GCL faces greater challenges in defending against adversarial attacks. Furthermore, there has been limited research on enhancing the robustness of GCL. To thoroughly explore the failure of GCL on the poisoned graphs, we investigate the detrimental effects of graph structural attacks against the GCL framework. We discover that, in addition to the conventional observation that graph structural attacks tend to connect dissimilar node pairs, these attacks also diminish the mutual information between the graph and its representations from an information-theoretical perspective, which is the cornerstone of the high-quality node embeddings for GCL. Motivated by this theoretical insight, we propose a robust graph contrastive learning framework with a learnable sanitation view that endeavors to sanitize the augmented graphs by restoring the diminished mutual information caused by the structural attacks. Additionally, we design a fully unsupervised tuning strategy to tune the hyperparameters without accessing the label information, which strictly coincides with the defender's knowledge. Extensive experiments demonstrate the effectiveness and efficiency of our proposed method compared to competitive baselines.